Establishing efficacy - without humans?

The decade following passage of FDAAA has been one of easing standards for drug approvals in the US, most notably with the advent of “breakthrough” designation created by FDASIA in 2012 and the 21st Century Cures Act in 2016.

Although, as of this writing, there is no nominee for FDA Commissioner, it appears to be safe to say that the current administration intends to accelerate the pace of deregulation, mostly through further lowering of approval requirements. In fact, some of the leading contenders for the position are on record as supporting a return to pre-Kefauver-Harris days, when drug efficacy was not even considered for approval.

Build a better mouse model, and pharma will
beat a path to your door - no laws needed.

In this context, it is at least refreshing to read a proposal to increase efficacy standards. This comes from two bioethicists at McGill University, who make the somewhat-startling case for a higher degree of efficacy evaluation before a drug begins any testing in humans.

We contend that a lack of emphasis on evidence for the efficacy of drug candidates is all too common in decisions about whether an experimental medicine can be tested in humans. We call for infrastructure, resources and better methods to rigorously evaluate the clinical promise of new interventions before testing them on humans for the first time.

The author propose some sort of centralized clearinghouse to evaluate efficacy more rigorously. It is unclear what they envision this new multispecialty review body’s standards for green-lighting a drug to enter human testing. Instead they propose three questions:

• What is the likelihood that the drug will prove clinically useful?
• Assume the drug works in humans. What is the likelihood of observing the preclinical results?
• Assume the drug does not work in humans. What is the likelihood of observing the preclinical results?

These seem like reasonable questions, I suppose – and are likely questions that are already being asked of preclinical data. They certainly do not rise to the level of providing a clear standard for regulatory approval, though perhaps it’s a reasonable place to start.

The most obvious counterargument here is one that the authors curiously don’t pick up on at all: if we had the ability to accurately (or even semiaccurately) predict efficacy preclinically, pharma sponsors would already be doing it. The comment notes: “More-thorough assessments of clinical potential before trials begin could lower failure rates and drug-development costs.” And it’s hard not to agree: every pharmaceutical company would love to have even an incrementally-better sense of whether their early pipeline drugs will be shown to work as hoped.

The authors note

Commercial interests cannot be trusted to ensure that human trials are launched only when the case for clinical potential is robust. We believe that many FIH studies are launched on the basis of flimsy, underscrutinized evidence.

However, they do not produce any evidence that industry is in any way deliberately underperforming their preclinical work, merely that preclinical efficacy is often difficult to reproduce and is poorly correlated with drug performance in humans.

Pharmaceutical companies have many times more candidate compounds than they can possibly afford to put into clinical trials. Figuring out how to lower failure rates – or at least the total cost of failure - is a prominent industry obsession, and efficacy remains the largest source of late-stage trial failure. This quest to “fail faster” has resulted in larger and more expensive phase 2 trials, and even to increased efficacy testing in some phase 1 trials. And we do this not because of regulatory pressure, but because of hopes that these efforts will save overall costs. So it seems beyond probable that companies would immediately invest more in preclinical efficacy testing, if such testing could be shown to have any real predictive power. But generally speaking, it does not.

As a general rule, we don’t need regulations that are firmly aligned with market incentives, we need regulations if and when we think those incentives might run counter to the general good. In this case, there are already incredibly strong market incentives to improve preclinical assessments. Where companies have attempted to do something with limited success, it would seem quixotic to think that regulatory fiat will accomplish more.

(One further point. The authors try to link the need for preclinical efficacy testing to the 2016 Bial tragedy. This seems incredibly tenuous: the authors speculate that perhaps trial participants would not have been harmed and killed if Bial had been required to produce more evidence of BIA102474’s clinical efficacy before embarking on their phase 1 trials. But that would have been entirely coincidental in this case: if the drug had in fact more evidence of therapeutic promise, the tragedy still would have happened, because it had nothing at all to do with the drug’s efficacy.

This is to some extent a minor nitpick, since the argument in favor of earlier efficacy testing does not depend on a link to Bial. However, I bring it up because a) the authors dedicate the first four paragraphs of their comment to the link, and b) there appears to be a minor trend of using the death and injuries of that trial to justify an array of otherwise-unrelated initiatives. This seems like a trend we should discourage.)

[Update 2/23: I posted this last night, not realizing that only a few hours earlier, John LaMattina had published on this same article. His take is similar to mine, in that he is suspicious of the idea that pharmaceutical companies would knowingly push ineffective drugs up their pipeline.]

Kimmelman, J., & Federico, C. (2017). Consider drug efficacy before first-in-human trials Nature, 542 (7639), 25-27 DOI: 10.1038/542025a

Jerry Matczak

Jerry Matczak passed away suddenly last Thursday at the much-too-young age of 54.

I can say, without exaggeration, that Jerry embodied pretty much everything I aspire to be in my professional life. The MedCityNews headline called him a “social media guru”, but in reality he was temperamentally the exact opposite of a "guru":

He was constantly curious; it seemed that every conversation I had with him was composed mainly of questions. Many of us try to be “listen first, talk second” types, but Jerry was a “listen first, ask questions, listen some more, then talk” type.

He also never stopped trying to figure out how to improve whatever he was working on. He participated in a lot of pilot projects, which means he was a part of a lot of projects that didn’t meet their objectives – but I never witnessed Jerry being the least bit negative or frustrated. Every project was just another opportunity to learn more.

Mostly, though, Jerry was remarkable in his ability to connect with patients, even patients who were deeply distrustful of his employer and industry. If nothing else, I hope you read the words of two such patients, coming from very different places, with remarkably similar reactions to Jerry:

• What Would Jerry Do? (by ALS Advocacy)
• Patients, Pharma, Partners (by AfternoonNapper)

Jerry, thank you for your service and your example. I carry it with me.

Angela @radclipatra & I rocking our #WalkingGallery of Healthcare jackets. Learn more https://t.co/uFZHW81Cts #SCOPE2017 @ReginaHolliday pic.twitter.com/6kamYOW2VZ

— Jerry Matczak (@gmatczak) January 25, 2017

The first paid research subject in written history?

On this date 349 years ago, Samuel Pepys relates in his famous diary a remarkable story about an upcoming medical experiment. As far as I can tell, this is the first written description of a paid research subject.

According to his account, the man (who he describes as “a little frantic”) was to be paid to undergo a blood transfusion from a sheep. It was hypothesized that the blood of this calm and docile animal would help to calm the man.

Some interesting things to note about this experiment:

• Equipoise. There is explicit disagreement about what effect the experimental treatment will have: according to Pepys, "some think it may have a good effect upon him as a frantic man by cooling his blood, others that it will not have any effect at all".
• Results published. An account of the experiment was published just two weeks later in the journal Philosophical Transactions. 
• Medical Privacy. In this subsequent write-up, the research subject is identified as Arthur Coga, a former Cambridge divinity student. According to at least one account, being publicly identified had a bad effect on Coga, as people who had heard of him allegedly succeeded in getting him to spend his stipend on drink (though no sources are provided to confirm this story).
• Patient Reported Outcome. Coga was apparently chosen because, although mentally ill, he was still considered educated enough to give an accurate description of the treatment effect. 

Depending on your perspective, this may also be a very early account of the placebo effect, or a classic case of ignoring the patient’s experience. Because even though his report was positive, the clinicians remained skeptical. From the journal article:

The Man after this operation, as well as in it, found himself very well, and hath given in his own Narrative under his own hand, enlarging more upon the benefit, he thinks, he hath received by it, than we think fit to own as yet.

…and in fact, a subsequent diary entry from Pepys mentions meeting Coga, with similarly mixed impressions: “he finds himself much better since, and as a new man, but he is cracked a little in his head”.

The amount Coga was paid for his participation? Twenty shillings – at the time, that was exactly one Guinea.

[Image credit: Wellcome Images]

Will Your Family Make You a Better Trial Participant?

It is becoming increasing accepted within the research community that patient engagement leads to a host of positive outcomes – most importantly (at least practically speaking) improved clinical trial recruitment and retention.

But while we can all agree that "patient engagement is good" in a highly general sense, we don't have much consensus on what the implications of that idea might be. There is precious little hard evidence about how to either attract engaged patients, or how we might effectively turn "regular patients" into "engaged patients".

That latter point - that we could improve trial enrollment and completion rates by converting the (very large) pool of less-engaged patient - is a central tenet of the mHealth movement in clinical trials. Since technology can now accompany us almost anywhere, it would seem that we have an unprecedented opportunity to reach out and connect with current and potential trial participants.

However, there are signs that this promised revolution in patient engagement hasn't come about. From the decline of new apps being downloaded to the startlingly high rate of people abandoning their wearable health devices, there's a growing body of evidence suggesting that we aren't in fact making very good progress towards increasing engagement. We appear to have underestimated the inertia of the disengaged patient.

So what can we do? We know people like their technology, but if they're not using it to engage with their healthcare decisions, we're no better off as a result.

Daniel Calvert, in a recent blog post at Parallel 6 offers an intriguing solution: he suggests we go beyond the patient and engage their wider group of loved ones. By engaging what Calvert calls the Support Circle - those people most likely to "encourage the health and well being of that patient as they undergo a difficult period of their life" - trial teams will find themselves with a more supported, and therefore more engaged, participant, with corresponding benefits to enrollment and retention. 

Calvert outlines a number of potential mechanisms to get spouses, children, and other loved ones involved in the trial process:

During the consent process the patient can invite their support team in with them. A mobile application can be put on their phones enabling encouraging messages, emails, and texts to be sent. Loved ones can see if their companion or family member did indeed take today’s medication or make last Monday’s appointment. Gamification offers badges or pop-ups: “Two months of consecutive appointments attended” or “perfect eDiary log!” Loved ones can see those notifications, like/comment, and constantly encourage the patients. 

Supporting materials can also be included in the Support Circle application. There are a host of unknown terms to patients and their team. Glossaries, videos, FAQs, contact now, and so much more can be made available at their fingertips.

I have to admit I'm fascinated by Calvert's idea. I want him to be right: the picture of supportive, encouraging, loving spouses and children standing by to help a patient get through a clinical trial is an attractive one. So is the idea that they're just waiting for us to include them - all we need to do is a bit of digital communication with them to get them fully on board as members of the study team.

The problem, however, remains: we have absolutely no evidence that this approach will work. There is no data showing that it is superior to other approaches to engage trial patients.

(In fact, we may even have some indirect evidence that it may hinder enrollment: in trials that require active caregiver participation, such as those in Alzheimer's Disease, caregivers are believed to often contribute to the barriers to patient enrollment).

Calvert's idea is a good one, and it's worthy of consideration. More importantly, it's worthy of being rigorously tested against other recruitment and retention approaches. We have a lot of cool new technologies, and even more great ideas - we're not lacking for those. What we're lacking is hard data showing us how these things perform. What we especially need is comparative data showing how new tactics work relative to other approaches.

Over 5 years ago, I wrote a blog post bemoaning the sloppy approaches we take in trial recruitment - a fact made all the more painfully ironic by the massive intellectual rigor of the trials themselves. I'm not at all sure that we've made any real progress in those 5 years.

In my next post, I'll outline what I believe are some of the critical steps we need to take to improve the current situation, and start bringing some solid evidence to the table along with our ideas.

[Photo credit: Flikr user Matthew G, "Love (of technology)"]

Waiver of Informed Consent - proposed changes in the 21st Century Cures Act

Adam Feuerstein points out - and expresses considerable alarm over - an overlooked clause in the 21^st Century Cures Act:

Waive informed consent requirement for clinical trials?!?! Unbelievable but true if #Path2Cures becomes law. pic.twitter.com/dqmWPpxPdE

— Adam Feuerstein (@adamfeuerstein) July 14, 2015

In another tweet, he suggests that the act will "decimate" informed consent in drug trials. Subsequent responses and retweets  did nothing to clarify the situation, and if anything tended to spread, rather than address, Feuerstein's confusion.

Below is a quick recap of the current regulatory context and a real-life example of where the new wording may be helpful. In short, though, I think it's safe to say:

1. Waiving informed consent is not new; it's already permitted under current regs
2. The standards for obtaining a waiver of consent are stringent
3. They may, in fact, be too stringent in a small number of situations
4. The act may, in fact, be helpful in those situations
5. Feuerstein may, in fact, need to chill out a little bit

(For the purposes of this discussion, I’m talking about drug trials, but I believe the device trial situation is parallel.)

Section 505(i) - the section this act proposes to amend - instructs the Secretary of Health and Human Services to propagate rules regarding clinical research. Subsection 4 addresses informed consent:

…the manufacturer, or the sponsor of the investigation, require[e] that experts using such drugs for investigational purposes certify to such manufacturer or sponsor that they will inform any human beings to whom such drugs, or any controls used in connection therewith, are being administered, or their representatives, that such drugs are being used for investigational purposes and will obtain the consent of such human beings or their representatives, except where it is not feasible or it is contrary to the best interests of such human beings.

[emphasis  mine]

Note that this section already recognizes situations where informed consent may be waived for practical or ethical reasons.

These rules were in fact promulgated under 45 CFR part 46, section 116. The relevant bit – as far as this conversation goes – regards circumstances under which informed consent might be fully or partially waived. Specifically, there are 4 criteria, all of which need to be met:

 (1) The research involves no more than minimal risk to the subjects;
 (2) The waiver or alteration will not adversely affect the rights and welfare of the subjects;
 (3) The research could not practicably be carried out without the waiver or alteration; and
 (4) Whenever appropriate, the subjects will be provided with additional pertinent information after participation.

In practice, this is an especially difficult set of criteria to meet for most studies. Criterion (1) rules out most “conventional” clinical trials, because the hallmarks of those trials (use of an investigational medicine, randomization of treatment, blinding of treatment allocation) are all deemed to be more than “minimal risk”. That leaves observational studies – but even many of these cannot clear the bar of criterion (3).

That word “practicably” is a doozy.

Here’s an all-too-real example from recent personal experience. A drug manufacturer wants to understand physicians’ rationales for performing a certain procedure. It seems – but there is little hard data – that a lot of physicians do not strictly follow guidelines on when to perform the procedure. So we devise a study: whenever the procedure is performed, we ask the physician to complete a quick form categorizing why they made their decision. We also ask him or her to transcribe a few pieces of data from the patient chart.

Even though the patients aren’t personally identifiable, the collection of medical data qualifies this as a clinical trial.

It’s a minimal risk trial, definitely: the trial doesn’t dictate at all what the doctor should do, it just asks him or her to record what they did and why, and supply a bit of medical context for the decision. All told, we estimated 15 minutes of physician time to complete the form.

The IRB monitoring the trial, however, denied our request for a waiver of informed consent, since it was “practicable” (not easy, but possible) to obtain informed consent from the patient.  Informed consent – even with a slimmed-down form – was going to take a minimum of 30 minutes, so the length of the physician’s involvement tripled. In addition, many physicians opted out of the trial because they felt that the informed consent process added unnecessary anxiety and alarm for their patients, and provided no corresponding benefit.

The end result was not surprising: the budget for the trial more than doubled, and enrollment was far below expectations.

Which leads to two questions:

1.       Did the informed consent appreciably help a single patient in the trial? Very arguably, no. Consenting to being “in” the trial made zero difference in the patients’ care, added time to their stay in the clinic, and possibly added to their anxiety.

2.       Was less knowledge collected as a result? Absolutely, yes. The sponsor could have run two studies for the same cost. Instead, they ultimately reduced the power of the trial in order to cut losses.

Bottom line, it appears that the modifications proposed in the 21^st Century Cures Act really only targets trials like the one in the example. The language clearly retains criteria 1 and 2 of the current HHS regs, which are the most important from a patient safety perspective, but cuts down the “practicability” requirement, potentially permitting high quality studies to be run with less time and cost.

Ultimately, it looks like a very small, but positive, change to the current rules.

The rest of the act appears to be a mash-up of some very good and some very bad (or at least not fully thought out) ideas. However, this clause should not be cause for alarm.