In the internet there is a big community of specialists involved in drug R&D and at LinkedIn there are quite some interesting discussions related to topics worth while considering. The question which haunts many drug developers is: Why is it that so many drugs flunk after phase II trials? That question is also immensly important for the development of drugs in the field of chronic pain. But it is a generic question. 

The starting question I put in was: "Why so many drugs fail in phase II? After having reviewed many pipelines in due dilligence analysis for big life science investors, I come to the conclusion that within the preclinical research phase some really big flaws are burried, reason why companies are losing too much funding during phase II, as too many drugs are ineffective. The screening batteries in the preclinical phase should be managed more professionally, using different models and really good dose finding and letting result be discussed by outsiders. Otherwise, there is too much of a my baby energy involbed. I am curious to find out your opinion about this very relevant topic." The LinkedIngroup was gathered under the header: Professionals in Pharmaceutical Industry and Biotech Industry.

Why so many drugs fail in phase II ?

Statement (me):

After having reviewed many pipelines in due dilligence analysis for big life science investors, I come to the conclusion that within the preclinical research phase some really big flaws are burried, reason why companies are losing too much funding during phase II, as too many drugs are ineffective. The screening batteries in the preclinical phase should be managed more professionally, using different models and really good dose finding and letting result be discussed by outsiders. Otherwise, there is too much of a my baby energy involbed. I am curious to find out your opinion about this very relevant topic.

Phase II failure and belief systems 

Comment 1 (Binhui Ben Ni, Experienced pharmaceuticals professional)

I do not dis-agree with you, However, translation of drug efficacy from an animal model to human disease is the key to be blamed on in addition to the heterogeneity in the pharmacogenetics of the tested population. 

Comment 2 (Bill Shannon,Associate Professor of Biostatistics in Medicine, Washington University, and Founder and President, BioRankings, LLC)

I believe there is a willingness to push forward with a product into Phase II based on the believe it will work and not on scientific and statistical evidence.

I have done statistical due diligence on Phase II and III trials and am sometimes amazed that any investor would back these drugs after the Phase I trial -- but a good sales pitch can sometimes override evidence. It does not surprise me there is a high failure rate in Phase II. 

Phase II failure and flaws in preclinical models 

Comment 3 (Bola Akinlade, MD. MBA, FACP Medical Director at Astellas Pharma)

I agree with the comment about the failure of pre-clinical models to accurately predict a drug's efficacy in P2 trials.

The literature is replete with many compounds that showed great activity in supposedly relevant pre-clinical models but then failed pilot clinical studies to demonstrate efficacy.

Most P1 studies are not designed to have an efficacy component and once safety in single and multiple ascending doses is demonstrated, the next phase - P2 then reveals the whether a drug truly has effiacy potential.

Until more robust and predictive pre-clinical models for human disease are developed, there would continue to be high failure rates in P2. The sliver lining that comes along with this finding is that it is better to fail in P2 than to rely on a Type 1 errror and proceed to more expensive P3 studies. 

Comment 4 (Oded VainasGroup Leader, Bio-Simulations at Optimata)

I believe that a more rational approach should be taken when translating data from pre-clinical and phase I trials to phase II.

This approach should include the use of modeling and simulations, either mathematical or statistical, and it seems that the pharmaceutical companies are starting to increase their use with such models. Of course there is no better model to predict the success in human trials than the human itself, however, models can supply execellent insights and decrease the huge money spending on trial and error. 

Comment 5 (me)

What Bill described and has seen, I have seen too. A agree completely. And Bola's silver line is actually a bit less silver and more black in my eyes...The in house animal models are also cherished too much...and how many negative findings go down the lab drain before entering a preclinical report...

I think the preclinical scientists should have the liberty to play and fool around a bit, getting to know a drug and its putative indications, strength and weaknesses, but once a serious start has been made, we need to bring in more critical mind that has been done so far.

Oded stated about modelling..definitely an interesting tool for pharmacokinetics etc...but its use is somewhat restricted, or would Oded know of a crisp example were we won significant time using a model? 

Phase II failure and biomarkers 

Comment 6 (Ronald NortonFounder at GCS - Global Clinical Research Solutions, LLC)

The discovery and use of relevant biomarkers has been the subject of much discussion since I was with large pharma in the 1980's. The good news is that we are really only now starting to have proven clinical utility applying in early phase trials. Add, at least in some therapeutic areas currently, the application of genetic markers, and one begins to the the opportunity for preclinical assessment or PD as well as PK. 

Comment 7 (Bill Shannon,Associate Professor of Biostatistics in Medicine, Washington University, and Founder and President, BioRankings, LLC)

When talking about biomarkers or genetic elements (e.g., SNPs, copy number variants) the problem with successful results in Phase I and failure in Phase II is even more exaggerated.

When scanning thousands or hundreds of thousands of markers (as is now done routinely) the likelihood of false positives is certain. I am not sure this is understood at the lab/clinical/management levels. When false positive results are pushed into Phase II trials they will most likely fail.

The only way to protect against false positives is to run confirmatory experiments -- no statistical/computational technique exists to decide if a marker is a false or true positive. (Technically this is known as the 'large P small N' problem where there are many more variables, P, than samples, N. It's impact on false positive rates is generally ignored, and probably more often not known by company decision makers.) 

Comment 8 (Ronald NortonFounder at GCS - Global Clinical Research Solutions, LLC)

Certainly, most of the now commonly applied screening methods in human trials went through the same routine. Broad application as industry learns what is clinically relevant. This is no surprise. Useful markers come to the forefront and less useful weeded out.

Today one would not argue against the utility of metabolic makers for assessment of PK. However, in the 70's and 80's it was less well accepted. The use of TPMT to set dosage for specific metabolic types is another example. Alzheimers - APO E. Oncology has a number of specifics as well. 

Comment 9 (VIRGILIO VINAS Clinical Studies Manager (research-regulatory FDA-EMEA)- Medical Affairs)

New drug evaluation development hasn't be update in applications like been in discoveries biomarkers applications is one of the tools hasn't be incorporate, more evaluation in nuclear cell activity and drug effect need to be an important point in the drug evaluation more than PK-PD. 

Comment 10 (Joga Gobburu Director, Division of Pharmacometrics, Office of Clinical Pharmacology at US FDA)

If we compare the producitivity (number of molecules surviving each phase) - PK scientists have improved their bridging capabilities across the phases. However, the uncertainty in translation of effectiveness/safety measures has not improved (Kola, Landis article).

Some publications below from our Division support Oded's point in concept. However, these publications did not include analyses of animal data. Having said that - it is even more troublesome to note that 50% registration trials fail, several for predictable reasons based on earlier trials. These points are discussed in some of the publications.

There are, I believe, non-scientific reasons too for this low productivity (e.g.: reward system).

1. Bhattaram VA et al. Impact of pharmacometrics on drug approval and labeling decisions: a survey of 42 new drug applications. AAPS J 2005;7:E503-E512.

2. Bhattaram VA et al. Impact of pharmacometric reviews on new drug approval and labeling decisions--a survey of 31 new drug applications submitted between 2005 and 2006. Clin Pharmacol Ther 2007; 81:213-221.

3. Wang Y et al. 2008. Leveraging prior quantitative knowledge to guide drug development decisions and regulatory science recommendations: impact of FDA pharmacometrics during 2004-2006. J Clin Pharmacol.;48:146-56

4. Gobburu JV and Lesko LJ. Quantitative disease, drug, and trial models. Annu Rev Pharmacol Toxicol 2009;49:291-301. 

Comment 11 (Carol GloffPrincipal and Owner at Carol Gloff & Associates)

I agree that sometimes drugs are taken into Phase 2 efficacy clinical trials when the nonclinical data do not warrant this. Many other times, however, there are either no animal models of the human disease to be treated, or the available animal models do not correlate well with the human disease and effect of treatments.

In addition, the reason that we do human studies is to determine if the drug (or biologic) is safe and effective. It seems unrealistic to me to expect that even the majority of drugs will demonstrate safety and efficacy in initial efficacy (Phase 2) studies.

If we could accurately predict safety and efficacy based on nonclinical studies, drug development would be a very different process.

I teach a course in regulatory affairs and each year I have at least one student who thinks that we should no longer do animal safety studies because "cells in culture and computers can tell us everthing we need to know about a drug's efficacy and safety." Too bad this is not the case! And, even whole animals don't predict what will happen in humans all that well, as I and others have discussed already.

So, do some drugs go into Phase 2 clinical trials when they shouldn't? Yes. However, in my opinion it is reasonable to expect that many drugs tested in human Phase 2 efficacy studies will fail. 

Phase II failure and pressure sharemarkets

Comment 12 (Rai KarklinsRai Karklins Pharmaceutical/Biotech/Business Consulting)

There are the pressures of "its my baby", "group think" and pressure from sharemarkets.

Taking small Biotechs with VC funding and milestone payment there big pressure for results. Such pressure does lead to suboptimum decision making, scientists may see what they want to see in the data, may exclude data using sophistry.

Common excuses for failed P1 or preclin are "wrong model", incorrect or inappropriate marker", some assay screw up etc etc.

The argument that Biotech is very complex and a bit of a black box can be used ad inifinitum to justify proceeding if you think about it.

As a Quality professional I can tell you an out of specification result in an FDA regulated lab get much more rigour and scrutiny then an out of expectation result in R & D; some may say for obvious reasons. But- I ask you think about that, the future of a whole research program and 100s of millions of $ let alone opportunity cost of not starting another project.

Some things that R & D departments do in labs dont come close to a fully cGMP QC lab, and of course they shouldnt be the same but as far as validity and reproducability they should meet good science criteria. It is a misnomer to think that pharma R & D do not need quality systems, I worked for years in petrochemicals and the R & D there had higher standards then some pharma labs I have audited (due to safety reasons alone!). CEOs of biotechs are under huge pressure to report on meeting development milestones, this pressure transfers to R & D and can make people take decisions they should not.

When one thinks about it the same conflict of interest between the Quality unit and operations exists in such a situation and relationship between the CEO and head fo R & D but its not regulated to avoid conflict of interests. 

Phase II and predictability 

Comment 13 (Oded Vainas Group Leader, Bio-Simulations at Optimata)

Carol Gloff stressed out previously three important points in her opinion:

1. Animal models don't predict what will happen in humans

2. In vitro and computers will not necessarily predict efficacy and toxicity

3. It is reasonable to expect that many drugs tested in human Phase 2 efficacy studies will fail

Well, I think that we can improve this by at least thinking of involving modeling and simulation teams during the whole project. The integration should be at each step of development and the approach is "learn and improve":

1. We check efficacy, toxicity and PK in animals --> predicting the three with scaling models to human--> performing phase I

2. We improve our models with phase I data and predict phase II --> improve models based on phase II data

3. We predict phase III --> perform it and improve models. During all steps it is important to take into account the modeling for PK, efficacy and toxicity, and NOT to rely only on PK. It is better to use semi-mechanistic modeling to describe the drug's MOA and not to rely only on data-driven models. This may also serve as a bridge between the pharmacometricians and the clinical/biology experts, that sometimes are reluctant to use a "black box".

We will improve our understanding of the drug, its efficacy and toxicity dynamics, and by utilizing simulations also calibrate the desired treatment protocol by taking into account PK, toxicity and efficacy In addition, we should expand our use with adaptive clinical trials, and to the best of my knowledge the FDA supports it.

This is exactly the place where Pharamcometrics can help: suggesting change in the regimen, deciding about the number of patients required, etc The problem is that usually modeling and simulation teams are approached only in times of disaster....

A naive question - I do not completely understand why it is so hard to include in the project teams a Pharmacometrician? I'm glad that Joga Gobburu posted previously examples supporting the involvement of modeling in the different phases of clinical trials and their importance. These papers are very important ( later today I will post others).

Comment 14 (me)

I am happy to see this is such a hot topic. The reason I put this in, is that I believe we are not very good in a crisp transition from preclinical to clinical. This is part due to scientific issues, such as being brought up above, for instance by Bill (stats) and Carol (no good animal models). And of couse there are subsolutions, such as the biosimulation thing. I do agree that can all assist in diminishing the chances of failure in phase II, and I consider these (PK, biosimulation and stats) as normal to rational drug development.

It is the key issue as described by Ray that I am worried about. The my baby thing, the pressure to fill up pipelines for shareholders value and all that making drug development partly a hot balloon topic. This is definitely also part of the greater picture why banks go broke, and contries nowadays. To push forward by presure of shareholders baby drugs in your own pipeline, based on inhouse animal models and not having the perspective of a scientist only, is in my opinion key to many of the failed and aborted drugs after phase II.

For instance idebenone in Freiedreich's disease. There were surrogate markers involved, but they were not appropriate, there is a formulation, but with lousy kinetics, and a clear dose finding was troublesome, and last but not least, the clinical read out was never validated. Now this was only an orphan drug...

There are a multitude of cases, and I have seen many in the field of CNS, were drugs have been taken up in phase II which shouldn't. Some important lessons, as I reviewed what I saw:

1. Benchmark preclincical findings in 1-2 models in house with 1-2 models in an academic lab somewhere

2. You need to find crisp dose-response relations, in different models!

3. You need to look into the dose! High doses in animal models might work, but in man we step down to suboptimal dose-ranges due to safety issues.

4. PK data in preclinical work is too flimsy. Some of my thoughts. Nice to see so many skilled input in this topic! 

Comment 15 (Steve Barton Independent Toxicology Consultant at BarTox Consulting Ltd)

A comment from a toxicologist: given that so many drugs fail in Phase 2, plan to defer (yes, I know it keeps me out of work) the toxicology needed for Phase 3 trials until at least there is initial indication of success in Phase 2. Alternatively, if delaying those toxicology studies would unduly delay Phase 3 studies, at least be aware that the studies may end up as money wasted. 

Comment 16 (me)

As this topic in my perception is adressing one of the Key Factors for Success and as there are already quite some valuable remarks made above, I intend, if everyone agrees, to put this whole discussion in a structured way in my blog/website somewhere in the near future. Including a commentary section. So that we do not loose the input of the great variety of people, who all bring into this topic their own ideas! I will let you all know via this medium were to surf to pick up the entire discussion in a structured way!  

Phase II, portfolio management and project management

Comment 17 (Laura BarrowWorldWide Head, Standard Operating Procedures at Pfizer Pharmaceuticals)

I'd like to bring the perspective of Project and Portfolio Management to the fore. There are few companies that have and maintain discipline around the criteria for advancing compounds from phase to phase. No matter what the data, simulations, models, and PK tell us they will use their "gut" to advance compounds that are near and dear to their heart. We can do all the pre-Phase II work we want to increase chances of success, but unless that data is heeded things will not change.

It differs also around whether you are investigating novel targets or a different approach to a known target. Novel targets will fail more often by nature but it is critical to determine that it is the target that has actually failed and not the specific compound failing to hit the target. I have seen the trend to actually do less pre Phase II (not of course eliminating the standard tox and PK screens) and get the compounds into man to see where we really are. PK in animals is at best an inconsistent predictor of PK in man, and the string above comments heavily on the lack of predictable pre clinical models for efficacy. I am puzzled by the suggestion to do more dose response relationship testing preclinically- I think the work in man would need to be done in any case. Again I have seen the emphasis not on the first compound which is used more as a probe for hitting the target, but on subsequent compounds of the same chemotype that are optimized for solubility, PK, etc. It depends on the number of novel targets you can afford to pursue and your R&D infrastrucutre.

In the end it also depends on the amount of risk the company is willing to sustain. When the pipeline is full and targets abound the criteria for advancement are stricter. In lean times there is a willingness to progress less robust compounds because- well that is all there is. Imagine what is left on the table during high number times and how much money is spent in lean times trying to make apples out of oranges. Not an easy business we are all in! 

Commeht 18 (Rai Karklins Rai Karklins Pharmaceutical/Biotech/Business Consulting)

Project management is also a key fact or lack of it. I have seen some companies embrace project management and others almost shun it as much as any kind of quality system. Thene there are the "soft " and "hard" models of project management which do you use depends on company culture. Too many meeting where what agreed at the meeting is not what happens when the meeting after the meeting occurs in the corrridors or labs. As they say, "if you dont know where you are going any road will get you there" and "projects become delayed by a year 1 day at a time".

Another thing I have seen is that the end result is not in mind- what is going to be on the patient leaflet or label claim? many times seems we make it up as we go along or R & D gets a nasty surpise at the end from sales and marketing. You would be surpised how often even the dose form (and technology to deliver it) is an afterthought. Having been involved in building facilities (which need to start before phase 3 even finished) I can attest to the word sterile coming up after basic design is completed due to a change of deliver. Project management is needed, science and no number of PhDs will actually deliver. 

Comment 19 (Bola Akinlade, MD. MBA, FACPMedical Director at Astellas Pharma)

I agree with most of the comments above. The bottom line is failure in P2 should be considered the norm rather than the exception for a variety of reasons as stated in the eloquent and insightful comments above. The challenge is how to decrease the high failure rate. It has got to be a combination of all of the above solutions. No single solution would suffice. However, a healthy dose of criticial thinking and refrain from group-think would help. Most of the comments have focused on innovative compounds and not on "me-too" compounds. There should be a lower rate of failure for "me too" compounds. Does anyone know what the literature says about this? 

Phase II failure and non-scientific factors 

Comment 20 (Joga GobburuDirector, Division of Pharmacometrics, Office of Clinical Pharmacology at US FDA)

Gordian, Singh, Zemmel and Elias (http://www.mckinsey.com/clientservice/pharmaceuticalsmedicalproducts/pdf/why_products_fail_in_phase_III_in_vivo_0406.pdf ) report that even for those compounds with established mechanism of action (MOA) and with objective endpoints (exhibit#3) the failure rate in late phase (not so-called phase 2) is 34% due to lack of effectiveness-differentiation. For similar endpoint, but new MOA, the failure rate if 63%. Hope this helps. This supports that there might be important non-scientific forces that influence decisions. 

Comment 21 (prof.dr. jan keppel hesselink Director R&D institute for neuropathic pain)

Yes Joga and colleagues, indeed there are other non-scientific factors..and more than we wish or that many of us are aware of. That is just the reason why I felt this discussion is quite worthwhile. And if these (dark, ;)) forces are present, we should make this more transparant.

One other issue related to the topic is the trend in our industry to focus on one pathogenetic mechanism only, related to one specific disease, the target indication. If you study the history of the last 20 years and put up together the various pathogenesis related to for instance stroke, starting with the lazaroids and the calcium antagonists...you will see a very interesting picture emerging. In drug development we try too much finding the proper disease for a drug already in the pipeline. Drugs looking for diseases was the titel of a PhD thesis years ago.

And we fold the pathogenesis in our presentation for CEO's and the board into that direction, don't we. It is like a survival of the fittest. If we in our own realm (departmet, lab) bring up a coherent or seemingly coherent picture, we get the bucks for further development. That is the game of being the product champion! It is about the sociology of scientific research. A not very popular branch in science, but a branch which is indeed important to look into. 

Comment 22 (Rai Karklins Rai Karklins Pharmaceutical/Biotech/Business Consulting) 

A question of mindset as well. The Research team should be also asking themselves "what are the torpedos waiting to sink this project" why should this one proceed and not another one? Scientists need to develop and grow such that they can also have a commercial business development mindset as well as scientific so they understand the concepts of value adding and opportunity costs. We always talked about the "funnel" of projects lots get assessed, very few proceed and even fewer succeed, so just like marketing its a numbers game.

Now- your scientists, project teams need to understand that: 1 Its actually not THEIR project 2 The company needs to proceed with the project when all things are considered is the BEST project (not always the most interesting one) 3 They need to know when to KILL a project and not let it become a cockroach that never dies (this is the most important skill in Pharma R & D and BD teams).

Often time though because the funnel is not full with project people are scared of losing their jobs and cling onto projects by whatever means, the one more experiment, the endless writing of reports etc etc I have shut down many projects and you can see the difference when the funnel is full and half full.

Like you said Jan, the sociology of research is not popular. Scientists by nature think they know everything and often dismiss the psychology/sociological side as poppycock. The issues with climate change are a start example of a giant R & D project and what can go wrong. It’s easy for very good scientists to talk about the intricate technical details but very difficult to deal with the behaviour side as its outside the comfort or expert zone- and believe it or not this can affect the outcome of projects. 

Comment 23 (Rai Karklins Rai Karklins Pharmaceutical/Biotech/Business Consulting)

Another old fashioned reason indirectly is the competition for resources within organisations. Some companies have an annual project budget cycle in which heads of project compete for scarce resources (along with other departments and other projects). If you have a project team with a project that is on thin ice, and it is not obvious what would or could replace it, then you may be tempted to may it look in better shape than it is. Depending on your presentation and marketing skills use of technical jargon etc you may convince the decision makers to keep it going.

This is an organisational design issue of how to keep the R & D machine working during feast or famine; don’t leave people to guess what will happen devise a system that people know, even if it means being split up and farmed out to other project teams. This point is important as it can provide a counter balance to teams becoming TOO strong and putting the team before all else; culture and values of self sacrifice and helping other project coupled to reward systems can help. Often some scientists for good reason are recognised within an organisation for their expertise; however this "expert and connections power" may provide a "right of way" for their projects and teams which is not deserved. Someone once said to me it is dangerous to ascribe honour and power for "past glory's" many years after it happened if there have been few since.

Sometimes these personalities have charisma and do the conference circuits but have lost the edge or not kept up to date, but back at the company no-one dare say anything simply because of the aura or charisma complex, especially if they are also narcissist (it can be a career limiting decision!). Some may simply brush this off as “politics”, that may be, but is that a reason for enduring phase 2-3 fails?

Coment 24 (me)

The point Ray makes, as well as did Joga and others is important to highlight. Ray also stipulates the narcissist issue. How many of us were not called upon to present a complicated project, a project which had to go through an important go-no go transition, for a circle of senior managers. The so called 'international development board', or whatever.

And how often pseudo-rational arguments, were fired from the side of some of these narcissitic managers, who liked to hear themselves talking, and had clearly alpha-male characteristics? By the sheer weight of their dominance, as well as by the group psychology, their voices heavily influenced projects and their future! Those were not scientific arguments, but power arguments, sometimes intermingled with a clear subjective liking of one dominant person for another group or another persons in the company, competing for resources... 

Comment 25 (Sarath KanekalPresident at WWW.TOX-CONSULTANT.COM)

I concur with Rai & Joga, having been in small biotechs for over 17 years. If executive management rewards are linked to solely reaching the clinical milestone (as opposed to delivering an effective drug candidate) a lot of unworthy ineffective candidates will be pushed into clinic-- seen this too often.

Decisions to push drugs into clinical trials are not always based on science especially in small pharma/biotechs-- and VCs need an exit! Fortunately for the patients we have FDA to make sure at least these ineffective clinical trial drugs are safe enough.

There are quite a few drugs in clinical trials which will never achieve therapeutically effective concentrations in humans due to DMPK or tox issues, something quite easily predictable from preclinical studies. That said, there are also scientific reasons why preclinical models do not predict clinical efficacy well, especially in oncology. 

Phase II faiure and other reasons 

Comment 26 (David CavallaFounder, Numedicus Ltd) 

Apart from the normal reasons, there are some other factors that pertain to current strategies which enhance, rather than reduce attrition rates:

-- new MoAs in an effort to enhance differentiation

-- biotechs progressing into Ph II with greater time and cost restraints than ever before

-- increasingly difficult therapeutic targets Joga is right that me-too approaches can give rise to non-scientific reasons (aka commercial) for attrition. What is needed are Goldilocks-type projects (neither too hot, nor too cold, but just right...).

They need to have sufficient differentiation to offer a competitive advantage, but at minimal risk. Such low-risk, high value approaches need to be carefully constructed, in order to avoid unnecessary failures while capturing enormous value.

See Nature Rev Drug Disc, 2009, 8, 849-853. Remember Nexium? 

Comment 27 (Nimisha SharmaResearch Scholar at Delhi University)

interesting discussion, i read an article which addressed the issue on novel drugs faling to pass phase III clinical trials, perhaps some of you might have already come across this article. Here's the pdf link: http://www.mckinsey.com/clientservice/pharmaceuticalsmedicalproducts/pdf/why_products_fail_in_phase_III_in_vivo_0406.pdf 

Comment 28 (Heather WebbDirector of Drug Safety. at Calistoga Pharmaceuticals)

The article posted by Joga Bobburu is supportive of the points made by several commenters. When a compound fails for reasons of insufficient efficacy it is not readily determined to what extent that is a failure of the intended target or a failure to explore and define the dose correctly.

This exploration best starts in the preclinical setting but is really optimally performed in the Phase 2 setting. If the preclinical dose response defines a optimal time course and extent of target modulation and the clinical safety margin and pharmacokinetics allow for recapitulating that time course and extent of target modulation then one can know that the compound failed for lack of efficacy.

However, if one can not assure that the intended target was modified in the intended way within the safety margin, how can one interpret the trial outcome? This seems to be failure due to insufficient trial design with insufficient supportive data. 

Comment 29 (Ralph CasaleDirector of Chemistry Operations at Ischemix) 

I think it may be useful to look at who is failing in phase II. While I don't have data to back up the opinion, it is my impression that more and more phase II candidates are from the realm of start-ups, often with few options other than to try to demonstrate a hint of efficacy in people. This is almost certainly going to result in a higher rate of failures.

Ad to that, more endpoints being evaluated at earlier stages gives drugs 'more opportunity to fail' during proof of efficacy evaluations. While not exactly on topic to the question, the 2006 congressional budget office look at the industry's pharmacoeconomics addresses many related issues and is likely of interest to this audience (if they haven't seen it already).

http://www.cbo.gov/ftpdocs/76xx/doc7615/10-02-DrugR-D.pdf 

Comment 30 (me)

Nimisha refers to quite an interesting paper. One of the main reasons for failure mentioned in that paper is that the failing drugs are as good as placebo. This brings in a new methodological-philosophical issue. I have always found that when we design phase II or III trials, and we implement many scales, each week one, we thus create a lot of attention energy if I may call it that way between the patient and the doctor. In that way we are maximizing the placebo response.

If you look into the placebo response for instance on antidepressants, in clinical trials from the late 70s till now, an increasing placebo response can be seen. The longer the trial, the more visits, the more centres and the bigger the protocol...the higher the placebo response.

This brings in one of the major flaws I percieve nowadays in clinical trials: the tendency to bring in many centres, in many countries, to recruit many patients in a short period of time... Even if I, as consultant for the pharmaceutical industry, discourage the clinical project team to select so many centres (and thus creating great variability and low protocol adherence) higher management overrules often due to pipeline management time deadlines... Ok we get the results earlier indeed,...but if the results are flat... 

Comment 31: Andy Potter

I have found this a very interesting topic to read about with many interesting perspectives. There is no doubt that the level of scrutiny in development and pre-clinical studies up to and including P1 is a factor in the potential failure of P2, however the timing and nature of P2 is perhaps the reason why a disproportionate number of drugs fail during this phase. Conclusive scientific results are much more convincing than scientific ommissions.

Most succesful sales people worth their salt would avoid offering potentially damaging eroneous results when making a case for P2 investment. I am not suggesting an intention to be misleading but that the facts regarding the products potential is infused with the hope of its success. Perhaps the fact P2 is a common stumbling area should be considered more of a meaningful transition given that these potential drugs were always destined to eventually fail? On the other side of the coin how many of todays succesful drugs would have not made it to market without the backing of a good sales team and maybe more importantly the investors who are not averse to a taking an educated and well informed gamble? 

Comment 32: Heather W.

The push to fail early has been a strong motivator for increasing early ADME/PK screening. Perhaps the same should be asked of compounds in non-clinical efficacy studies. But the predictability of unproven targets in nonclinical setting is a poor predictor. Good dose range finding in the clinic with robust clinical markers of target modulation and disease endpoints are the best markers for P3 success. The idea of failing early puts strong pressures on the decision to go into P3 which has moved the failure rate back into P2. P2 was once simply a time to do dose exploration to define the dose for P3. Now we require it to provide evidence of efficacy or even superiority. Thus, the failure rate in P2 has to do with many compounds coming into P2 with less than optimal as well as the higher bars set for passing into P3. Perhaps the question should be 'What is special about the compounds that succeed in P2?'. Is it the corporate organizational support? Is it good marketing by the project team? Is it the preclinical evidence going into P2? Is it good dose range exploration in P1/2? Is it the clinical design for P2 with robust markers of target and disease? Why do they succeed where others fail? 

Comment 33: Jean-Pierre M.

I will certainly not question the usefullness of animals models, however check how much care is taken into standardization of animals models as far as genetic background, age, gender, housing conditions and initial status of the disease state. The results obtained, altough fully valid and generally confirmed by strong statistical significance, poorly translate into the human species where genetic background is greatly diverse, age and social conditions vary, daily conditions are as diverse as patients, which is in complete opposition with the conditions of proof of principal animal studies. Overall, this high failure rate in Phase II might validate the personnalized medecine approach. So much for the blockbusters! My 2 cents

Comment 34: Jeffrey P.

Reasons for high failure rate: -Many NCEs target physiological systems whose regulation or relevance in humans are not well understood. - Preclinical models of efficacy have poor predictability because they were validated in a narrow way, based on drugs whose mechanism has been well established in humans 

Comment 35: Yulia O.

Do you think that the primary human cell models may prove to be a valuable tool for pre-clinical efficacy, MOA or toxicity testing? A wide variety of cryopreserved primary human cells became available in the past 2-3 years. Improved cryopreservation, cell culture techniques and use of 3-D matrices simplified development of predictive primary human cell screening assays. 

Comment 36: prof.dr. jan keppel hesselink

@ Yulia: definitely new models can help a lot, especially when used as screening tools in a logically build up battery. And if the tests compared to other tests provide coherent read outs. But the major issues are to be found in the animal models, and the PK and PD issues, the predictability, as Jeffrey pointed out, and the absence of a gold standard in the market place in indications like stroke, SAH, SCI, Brain Injury, Alzheimer, Neuropathic pain and the like. And in other indications we use gold standard and thus we find only me too's. Depression models validated with depressants will yield more of the same antidepressants. And that is what we all have seen, apart from the fact that we know meanwhile that nearly all antidepressants do not work in clinical mild to moderate depression at all....

Last week I discussed this topic with the CEO of Elan, at the Global Biotechynology Congress of Ernst and Young and LSP in Wassenaar, the Netherlands. In this discussion one major issue popped up: ownership. In big companies there are so many guidelines and restrictions, that there is only little space for ownership. He suggested to spin out from big companies groups of scientists and let them do their own thing under the umbrella from the big company, but without all the restrictions, all the board meetings and presentations and all the Kafka paper work. Give them more freedom! Sounded to me like a really good proposal. 

October 2010, Jan M. Keppel Hesselink, MD, PhD,

Professor molecular pharmacology