The Case against Randomized Trials is, Fittingly, Anecdotal

I have a lot of respect for Eric Topol, and am a huge fan of his ongoing work to bring new mobile technology to benefit patients.

The Trial of the Future

However, I am simply baffled by this short video he recently posted on his Medscape blog. In it, he argues against the continued use of randomized controlled trials (RCTs) to provide evidence for or against new drugs.

His argument for this is two anecdotes: one negative, one positive. The negative anecdote is about the recently approved drug for melanoma, Zelboraf:

Well, that's great if one can do [RCTs], but often we're talking about needing thousands, if not tens of thousands, of patients for these types of clinical trials. And things are changing so fast with respect to medicine and, for example, genomically guided interventions that it's going to become increasingly difficult to justify these very large clinical trials. 

For example, there was a drug trial for melanoma and the mutation of BRAF, which is the gene that is found in about 60% of people with malignant melanoma. When that trial was done, there was a placebo control, and there was a big ethical charge asking whether it is justifiable to have a body count. This was a matched drug for the biology underpinning metastatic melanoma, which is essentially a fatal condition within 1 year, and researchers were giving some individuals a placebo.

First and foremost, this is simply factually incorrect on a couple extremely important points.

1. Zelboraf was not approved based on any placebo-controlled trials. The phase 1 and phase 2 trials were both single-arm, open label studies. The only phase 3 trial run before FDA approval used dacarbazine in the comparator arm. In fact, of the 34 trials currently listed for Zelboraf on ClinicalTrials.gov, only one has a placebo control: it’s an adjuvant trial for patients whose melanoma has been completely resected, where no treatment may very well be the best option.
2. The Zelboraf trials are not an example of “needing thousands, if not tens of thousands, of patients” for approval. The phase 3 trial enrolled 675 patients. Even adding the phase 1 and 2 trials doesn’t get us to 1000 patients.

Correcting these details take a lot away from the power of this single drug to be a good example of why we should stop using “the sanctimonious [sic] randomized, placebo-controlled clinical trial”.

The second anecdote is about a novel Alzheimer’s Disease candidate:

A remarkable example of a trial of the future was announced in May. For this trial, the National Institutes of Health is working with [Banner Alzheimer's Institute] in Arizona, the University of Antioquia in Colombia, and Genentech to have a specific mutation studied in a large extended family living in the country of Colombia in South America. There is a family of 8000 individuals who have the so-called Paisa mutation, a presenilin gene mutation, which results in every member of this family developing dementia in their 40s. 

Researchers will be testing a drug that binds amyloid, a monoclonal antibody, in just 300 family members. They're not following these patients out to the point of where they get dementia. Instead, they are using surrogate markers to see whether or not the process of developing Alzheimer's can be blocked using this drug. This is an exciting way in which we can study treatments that can potentially prevent Alzheimer's in a very well-demarcated, very restricted population with a genetic defect, and then branch out to a much broader population of people who are at risk for Alzheimer's. These are the types of trials of the future. 

There are some additional disturbing factual errors here – the extended family numbers about 5,000, not 8,000. And estimates of the prevalence of the mutation within that family appear to vary from about one-third to one-half, so it’s simply wrong to state that “every member of this family” will develop dementia.

However, those errors are relatively minor, and are completely overshadowed by the massive irony that this is a randomized, placebo-controlled trial. Only 100 of the 300 trial participants will receive the active study drug, crenezumab. The other 200 will be on placebo.

And so, the “trial of the future” held up as a way to get us out of using randomized, placebo-controlled trials is actually a randomized, placebo-controlled trial itself. I hope you can understand why I’m completely baffled that Topol thinks this is evidence of anything.

Finally, I have to ask: how is this the trial of the future, anyway? It is a short-term study on a highly-selected patient population with a specific genetic profile, measuring surrogate markers to provide proof of concept for later, larger studies. Is it just me, or does that sound exactly like the early lovastatin trials of the mid-1980’s, which tested cholesterol reduction in a small population of patients with severe heterozygous familial hypercholesterolemia? Back to the Future, indeed.


[Image: time-travelling supercar courtesy of Flickr user JoshBerglund19.]

Accelerated Approvals are Too Fast, Except When They're Too Slow

A great article in Medscape reports on two unrelated articles on the FDA’s process for granting (and following up on) Accelerated Approvals of oncology drugs.

First, a very solid review of all oncology drugs approved through the accelerated process since 1992 is in the latest journal of the National Cancer Institute. The review, written by FDA personnel, is in general concerned with the slow pace of confirmatory Phase 3 trials – over a third (18 of 47) have not yet been completed, and even the ones that have completed have taken considerable time. The authors consider process changes and fines as viable means for the FDA to encourage timely completion.

Second, over at the New England Journal of Medicine, Dr Bruce Chabner has a perspective piece that looks at the flip side: he argues that some compounds should be considered even earlier for accelerated approval, using the example of Plexxikon’s much-heralded PLX4032, which showed an amazing 80% response rate in Metastatic Melanoma (albeit in a very small sample of 38 patients).

I would argue that we are just now starting to get enough experience to have a very good conversation about accelerated approval and how to improve it -- still, less than 50 data points (47 approved indications) means that we need to remind ourselves that we're still mostly in the land of anecdote. However, it may be time to ask: how much does delay truly cost us in terms of our overall health? What is the cost of delayed approval (how many patients may potentially suffer from lack of access), and correspondingly what is the cost of premature approval and/or delayed confirmation (how many patients are exposed to ineffective and toxic treatments)?

The good news, to me, is that we're finally starting to collect enough information to make a rational estimate of these questions.

Realistic Optimism in Clinical Trials

The concept of “unrealistic optimism” among clinical trial participants has gotten a fair bit of press lately, mostly due to a small study published in IRB: Ethics and Human Research. (I should stress the smallness of the study: it was a survey given to 72 blood cancer patients. This is worth noting in light of the slightly-bizarre Medscape headline that optimism “plagues” clinical trials.)

I was therefore happy to see this article reporting out of the Society for Surgical Oncology. In looking at breast cancer outcomes between surgical oncologists and general surgeons, the authors appear to have found that most of the beneficial outcomes among patients treated by surgical oncologist can be ascribed to clinical trial participation. Some major findings:

• 56% of patients treated by a surgical oncologist participated in a trial, versus only 7% of those treated by a general surgeon
• Clinical trial patients had significantly longer median follow-up than non-participants (44.6 months vs. 38.5 months)
• Most importantly, clinical trial patients had significantly improved overall survival at 5 years than non-participants (31% vs. 26%)

Of course, the study reported on in the IRB article did not compare non-trial participants’ attitudes, so these aren’t necessarily contradictory results. However, I suspect that the message of “clinical trial participation” entails “better follow-up” entails “improved outcomes” will not get the same eye-catching headline in Medscape. Which is a shame, since we already have enough negative press about clinical trials out there.