Pharma’s Cutting Edge

Consumers Union Petition

Petition for Better Reporting of Drug Side Effects
I support requiring all drug ads to include a 1-800 number and website so citizens can report drug side effects (petition docket 2008P-0012/CP1).

All too often, drug advertisements fail to present the benefits and risks of using prescription drugs in an accurate and balanced way. It is often the newest drugs that are the most heavily advertised, and it is these drugs whose side effects we know the least about.

The current system for collecting information about side effects catches only a fraction of actual cases. The recent law passed to require print drug ads to provide a 1-800 number and website (FDAAA — P.L. 110-85) is a step in the right direction, but should be extended to include TV ads which are viewed far more frequently and with a greater command of the viewing audience.

Increased reporting of adverse events will help in the earlier detection and better analysis of problems. All television ads should contain information on how patients can report side effects to the FDA.

As you can read above, Consumers Union is trying to improve surveillance of drug safety by promoting patient reporting of “side effects” in drug ads.  A representative of the organization had asked me to blog about the petition, so I am. 

I oppose their effort. Here’s why.

Probably around 1% of adverse reactions to drugs are reported via FDA’s Medwatch system.  Most reports come from healthcare professionals, based on my experience reviewing such reports in the 90’s.  That’s the way it should be, and we need to encourage more of it. 

On the other hand, encouraging more reporting af “side effects” by consumers might actually have its intended effect, which could completely break an already shaky surveillance system. 

Pasive surveillance of reported events can be an effective way of uncovering signals that a drug is associated with rare but potentially serious adverse effects (like hepatic toxicity) or with adverse effects that occur on a background of a common disease (like heart disease) at moderate relative risk (like the risk of heart attack with Vioxx, for instance). 

There are real challenges when interpreting data from such types of surveillance programs.  The most obvious challenge is that there is no concurrent control group for comparison to the drug-treated group (when there is a concurrent control group, we’re speaking of a form of surveillance that is different from the type I’m describing).  A less obvious challenge is separating signal from noise in the data.  Signals are so-called true-positive adverse event reports.  Noise comes from false-positive reports.  Safety surveillance systems strive to maximize signal and minimize noise in order to provide the best possible opportunity for making correct inferences from the data. 

When there is a low percentage of adverse events being reported overall, as in the current environment, the signal is low, and so the detrimental effects of any noise are amplified.  The situation will be improved by encouraging more adverse reports with a high chance of being true-positives and by discouraging reports with a high chance of being false-positives.  How do we do this?

We don’t do it by encouraging patients to fill out Medwatch forms or by calling 800 numbers.  We do it by making it much easier for doctors to capture and transmit adverse events that are possibly or probably related to drugs and by making it much easier for FDA to capture and analyze such reports.  The technology for facilitating such data capture and interchange already exists.  The technology on the doctor end is the electronic medical (health) record (EHR).  Interchange allowing upload of data to FDA can be facilitated using the HL7 data standard known as ICSR.  And on the back end, FDA already uses analytical tools that can make sense of the incoming data.  It is simply a matter of mandating use and providing funds for its implementation. 

Doctors can’t afford EHR for the most part, which is the main reason why its adoption has been so painfully slow.  In an EHR environment, a passive surveillance system can flourish by giving doctors easy, pain-free ways to report anonymized AE reports directly to FDA or other surveillance agencies (like CDC).  It can be as easy as adding a couple of tick boxes and a “send” button to a standard electronic patient encounter form (okay there’s more to it than that, but no so far as end-users can tell).

The US government has not done enough to encourage adoption of this technology, which would really improve drug safety surveillance.  The government should be buying EHR systems and associated training and support for every physician office until the technology achieves saturation.  By becoming the buyer, HHS can mandate that systems be sematically interoperable (be able to talk to each other and to FDA unambiguously).

The cost wouldn’t be as high as you might think.  According to the 2007 US economic census (2002 data), there are roughly 205,000 physician offices in the US.  Some of these offices (let’s say for argument-sake it’s 10%) already have EHR systems, but let’s figure that the government provides support for every office.  Because of discounts from competitive bids, the average cost per office will be much lower than it is today.  Let’s say it’s $50,000 per office on average.  That’s roughly 10.25 billion dollars to outfit, train, and support every physician office in the US with an EHR system and to guarantee that the systems will be interoperable with each other and with FDA.  That’s about what the US is spending each month in Iraq.

That’s what Consumers Union should be pushing for.  Adding to the noise of passive surveillance data will only make a bad situation worse.

PDUFA Reauthorization: Almost There

As Motley Fool and others are reporting, we can expect to see the 2007 bill reauthorizing the Precription Drug User Fee to emerge from conference committee this week, as FDA will give notices of layoffs to 2,000 employees on Sept. 21 without the legislation passing both houses and being signed into law by the president.  Conference committee sessions are closed to the public.  Congressional aides have been reporting that language providing a pathway for FDA approval of biosimilars aka biogenerics has been removed from consideration of the PDUFA legislation but could be debated as part of stand-alone legislation later this year.

I will report on the contents of the proposed PDUFA legislation as it is made public.  It’s fair to assume that morale and workflow are being disrupted within FDA while it awaits word on the bill.  It’s also fair to assume that such distractions are not good for sponsors. 

I read BusinessWeek.  It’s a quick general overview of what’s going on in business, the economy and investing; I can’t imagine that anyone thinks it’s more than that.  I certainly can’t imagine that anyone would use it to make individual stock picks.  Yet, for 10 years Gene Marcial has been providing his “insights” into individual stocks.  I view Marcial’s column the same way I view any investment newsletter–with great skepticism.  How does he select stocks for inclusion?  How does he select the analysts he quotes?  What ties doe does he have to any of these people or their companies (he doesn’t take stock positions in the companies he mentions). 

The reason I’m even discussing Marcial’s column is that he frequently discusses biotech companies.  He makes it pretty clear to anyone who’s been in or around the industry for a while that he knows this industry less than well–that is to say barely or perhaps not at all.  But that’s okay.  If he always chooses his sources like a chop house sommelier chooses the red to match your Kobe, he could still make wise calls and followers of his cobbled together advice might still make out.

But he’s no stock-picking somellier. 

What irks me about the column isn’t even so much that Marcial doesn’t know a thing about biotech but that BusinessWeek sees fit to tout his long-term performance.  Long-term performance?  If he’s a lousy biotech stock-picker, as I claim, why would BusinessWeek tout his long-term performance?  And there’s the rub.  It’s because BusinessWeek lies.  Well…not so much lies as distorts, in a way that inflates Marcials’ picking prowess.  Here’s how:  Marcial’s column is published online on a Thursday afternoon, after the market closes.  By the time anyone can act on the column it’s at least Friday (or Monday if it’s a 3-day weekend).  By the time the market opens on Friday, the “Marcial effect” has occurred and the stock price at open reflects it.  Not surprisingly, when Marcial touts a stock Thurday it frequently bumps Friday.  Not that any individual investor could take advantage of the bump, unless he/she knew that Marcial’s column was about to be published.  Following the bump, the market settles to reflect the stock’s underlying value.  The distortion from BusinessWeek comes from the price they use to reflect Marcial’s total returns.  They use as the basis the stock price at close on Thurday, not the price at open on Friday.  So, the 1-day return is from the close Thursday to the close Friday.

You can see the “Marcial effect” by reviewing Marcial’s 1-day returns compared with broad indices, as BusinessWeek has published in their August 13th issue.  You’ll see that Marcial beats the S&P 500, the DJIA, and the Russell 2000 by 3.8, 3.7, and 3.8%, respectively.  Folks, that’s not a sign of stock-picking prowess, that’s a sign of a direct influence on stock market valuations.  If Marcial really did pick stocks that beat the indices over a 10-year period as BusinessWeek claim, we would expect that his 1-day returns would closely approximate the indices’ 1-day returns.  That’s because a broad selection of stocks will never appreciate measured over any randomly selected 1-day period, because their combined underlying value will never increase significantly over a 1-day period.  So, BusinessWeek’s own published data shows that the “Marcial effect” contributed to nearly a 4% transient bump in the value of his picks.  Read the rest of this entry »

Thanks to Derek Lowe for bringing to my attention Arnold Relman’s long-winded answer in the New Republic (reg. required) to Richard Epstein’s book on the drug industry, Overdose, published last year.  I thought I’d spend far less space than Dr. Relman illuminating some of the ‘facts’ he uses in his arguments against Epstein and the industry.  I will also note out of fairness to Dr. Relman that I find myself in agreement with much of what he argues against.  I am in particular agreement with his views about the need to regulate the sale and marketing of drugs and the myth that drug prices are directly tied to R&D costs, as many of my prior posts will attest.  Relman’s words words are italicized; mine plain text.

Adding to the obscurity of these estimated research and development costs are a number of other inconvenient facts. First, most new drugs entering the market these days are not new molecular entities….

Industry critics frequently cite the decline of innovation since the mid-1990’s as evidence of the industry’s ills; they support this assertion with the suggestion that the decline of new molecular entity (NME ) approvals is symptomatic of a larger trend towards development of what are perjoratively termed ‘me-too’ drugs and away from ‘truly innovative’ drugs (Relman calls ‘me-too’ drugs an “inconvenient fact.”  While it may be more convenient to speak only of NMEs when discussing drug-industry innovation, follow-on drug products have important therapeutic and economic benefits that must be discussed in order to understand the total value of drug-industry innovation; too bad there are few within the industry who feel compelled to extoll their virtues).  Please refer to my article published a couple years ago in NRDD and available through the PGS website, which presents some of the information I will refer to.  The primary source of drug-approval information in the U.S. is provided by FDA and is freely available.  There is no question that productivity, as defined by the cost of R&D per dollar-value of approved drugs has been in decline.  The productivity decline appears much more pronounced when productive output is defined by the number of new molecular entities approved per dollar of R&D spent.  It is also true that most new drugs entering the market these days are not NMEs, and that the proportion of NDAs approved by the FDA that are NMEs has declined since 1990, as shown below (data derived from FDA tables).  NMEs make up a higher proportion of drugs when only NMEs and follow-on drug NDAs are counted (range 22%-88%), but the time trend is similar to what is shown.  So far, Relman and I are in agreement over the facts about NME approvals.

…and therefore they [follow-ons] are less costly to develop and to test than the new molecules for which cost estimates are made.

So much for our agreement.  I am aware of no published data that purport to show that costs to develop a follow-on product are lower than the costs to develop an NME, all else held equal.  Indeed, DiMasi and Paquette’s study on the economics of follow-on drug development, while not directly demonstrating the equivalency of costs of development for pioneering versus follow-on drugs, does support this proposition that follow-on drugs are no cheaper to develop than pioneers.  The authors eloquently demonstrate that the distinction between pioneering and follow-on drugs has become blurred by the quickening race to market.  Improvements to existing drugs are a different story.  It is reasonable to assume, for instance, that the cost of developing a new formulation of a previously approved active ingredient will be less costly than the original formulation.  If that is what Relman meant, it is not what he said. Read the rest of this entry »

I’ve been following along with the circus that is Ariad v Lilly for over a year now (see 2006 posts May 5, June 28, June 6), and, as I find it fascinating, I’m going to keep writing about it.  If patent decisions bore you, read no further.  The latest news is significant: Ariad has prevailed in a bench trial held in Massachusetts federal court.  The prior jury award, giving Ariad entitlement to 2.3% of revenues from sales of Xigris and Evista (valued at ~$65 million at the time of the award), stands for now.  Lilly has not yet commented on the Court’s decision.  At the same time Ariad was getting this bit of good news, they received word from the USPTO that the patent claims upheld by the Massachusetts Court were nullified for now.  Ariad will appeal the USPTO decision, which will take one to two years to be finally decided.  In the meantime, the award against Lilly will stand unless appealed.

I’m going to walk through the judge’s reasoning in this case.  If you read this and think you can successfully defend the logic of Judge Zobel, which I shall try my best to impugn, please do so, and more fool me.  The judge’s verbatims are indented.  The Court begins with a succinct summary of the case and his ruling:

Plaintiffs Ariad Pharmaceuticals, Inc., Massachusetts Institute of Technology, the  Whitehead Institute for Biomedical Research, and the President and Fellows of  Harvard College (collectively “Ariad”), owners and assignees of U.S. Patent No.  6,410,516 (“the ‘516 patent”), “Nuclear Factors Associated With Transcriptional  Regulation,” complain that defendant Eli Lilly & Co. (“Lilly”) infringed it. Following a  fourteen-day trial in April 2006, a jury found that the four asserted claims were valid  against anticipation, enablement and written description defenses, and that use of  Lilly’s Evista and Xigris products infringed the patent. The jury awarded plaintiffs  damages in excess of $65 million.  The parties agreed that certain additional defenses were to be tried to the court.  Lilly asserts that the ‘516 patent is invalid because it attempts to claim non-patentable  subject matter under 35 U.S.C. § 101. Even if the patent is valid, Lilly argues (1) that it  cannot be enforced because of inequitable conduct by plaintiffs during the prosecution  of the patent, or in the alternative; (2) that plaintiffs are estopped from recovering for  any infringement because they unreasonably delayed prosecution of the patent.  Following a second trial focused on these issues, I find that: (1) the four claims asserted are patentable; (2) Lilly has not proven inequitable conduct during patent  prosecution; and (3) Ariad did not unreasonably delay prosecution of the ‘516 patent.  Accordingly, the jury award stands.

The Court describes the extent of the claims allowed by the ‘516 patent:

The allowed claims broadly cover a method of inhibiting the expression of a  gene whose transcription is regulated by NF-κB in a eukaryotic cell.8 The only step  required to practice the broadest patented method is to “reduc[e] NF-κB activity in the cell such that the expression of said gene is inhibited.”9 No particular agent or substance need be used, nor any particular step(s) performed, to reduce NF-κB activity in order to practice the invention.

Skipping to specifics of Lilly’s assertion that the ‘516 claims aren’t patentable (because they attempt to claim inventorship of a natural phenomenon):

In Lilly’s view, the ‘516 patent claims subject matter not allowed under 35 U.S.C.  § 101 and is therefore invalid. Specifically, it contends that the claims encompass the NF-κB-IκB autoregulatory loop (the “Autoregulatory Loop”), a natural process in cells that operates to reduce the activity of NF-κB. Because natural phenomena are excluded from patentable subject matter, it argues that any ‘516 claims encompassing  the Autoregulatory Loop are invalid and cannot be enforced.  Ariad’s response is that the ‘516 patent does not claim a natural phenomenon because, inter alia, (1) the patent claims a process, subject matter specifically allowed  by statute; and (2) the Autoregulatory Loop is only a theory and has not been proven to  exist in human cells in vivo. (Docket # 398 ¶ 618). While not all processes are patentable, I find that Lilly has failed to show that the proposed model of the Autoregulatory Loop actually exists in nature and thus that a natural phenomenon is encompassed by the ‘516 patent’s claims.