Liquid Biopsy Continues to Disappoint

Confession: I have been burned by Guardant twice. Both times were my fault, the first on account of my wide-eyed idealism and the other because I didn’t stick to my guns when it counted, but this has seeded a personal and possibly misdirected venom toward the company.

It helps that the product sucks.

You generally avoid cutting into IBC, because it’s inherently inflammatory and there’s no reason to make it angry. My strategy is to tiptoe around it as quietly as possible. For this reason I’ve avoided a lot of biopsies, and I’m also resistant to the idea that you should re-biopsy patients at every progression event (this isn’t just an IBC objection; subjecting patients to a lung or a liver biopsy at anything resembling a regular basis seems cruel and dangerous – and for what?). Imagine my excitement when I learned I could ascertain the mysteries of my tumor genetics from a simple blood test!

I was living in Indiana at the time (don’t ever do that), so naturally I had to travel a great distance to find a physician who offered the Guardant test. There was minimal clinical data (single-site, I think, like a cohort of 15 patients), no reason to believe that anything useful would come of it, but I could avoid a biopsy. I had the blood draw. Three weeks later I got the results: no circulating tumor DNA identified, and thus no report. Well, shit. That was in 2015.

The Guardant test was offered to me a couple more times in the past two years; my policy was to decline. As documented previously, I was eager to try the new Foundation Medicine liquid biopsy; also a fail. No ctDNA. So I figured I didn’t have any circulating tumor DNA and went on living my life.

A couple months ago, I was back in my oncologist’s office, and Guardant came up again. No, I insisted. I just had the Foundation liquid biopsy; it didn’t work; my skin isn’t progressing; nothing’s changed, so no. I’m sure I said no at least twice. Then I proceeded to try to get myself on a boring local therapy study, which I ended up not starting; this is only relevant because as I was signing the three thousand pages of consent documents, pen warm in my hand, a pre-completed Guardant requisition form was at the bottom of the stack. I saw my name. The only thing missing was my signature. I grumbled, I said it was a waste, and then I signed it. I signed it!

You guys, I let you down. I let myself down. This is not what I stand for. I know better! I am still kicking myself over this. If I can’t say no three times in a row, what hope does the Resistance have against the vile corrupt institution of Big Oncology? Why did I DO that? I know this sales tactic; you get people saying yes, and they keep saying yes.

Note: everything you learn in a used car dealership will prepare you to be a cancer patient.

I subsequently found out that this time, the Guardant test reported a p53 mutation and a PI3K mutation. I simultaneously had a Foundation One tissue test, and it found a HER2 mutation and a PI3K mutation, and though the Guardant missed the only actionable mutation (HER2) and the p53 is an obvious false positive, it seemed inside the realm of possibility that the Guardant test sucked 2% less than in 2015, and it managed to correctly call my PI3K mutation. Not actionable, but maybe accurate? That’s something, right?

Except not. On Twitter today I saw report describing the wild discordance between Foundation’s tissue sequencing and the Guardant liquid biopsy. Actionability aside, let’s assume Foundation One is a reasonable standard for tissue sequencing. Whether or not treatment decisions should be based on the results and whether this improves outcomes is obviously up for debate (lots of positive case studies, but the larger studies just do not seem to deliver; I’m on the record as a believer in spite of this), I’ve never heard anyone dispute the accuracy of the Foundation One results. The concordance rate between Foundation One and Guardant 360 in this study was 8.6%. If tissue is the more reliable method of assessing tumor DNA, that means Guardant just … does not work. It’s not my anecdote. It doesn’t work.

The PI3K example in the paper jumped out at me. Of the 19 patients in the study, there were eight who had PI3K mutations per Foundation One. There were two who had PI3K mutations per Guardant. The concordance rate was zero, suggestive of false positives.

PI3K is a common mutation, with this study reporting that when 20,000 consecutive cases were analyzed (representing >40 cancer types), PI3K was mutated in 38% of cases. It’s an active area for drug development, and while we only really have Afinitor right now, there’s a whole wave of PI3K drugs coming in the next couple of years. Guardant is right to spit out a PI3K result, stopped clocks and all, but if we start treating patients based on these results … imagine the waste, and the threat to personalized medicine. I do think sequencing should inform treatment decisions and will improve patient outcomes, but not if idiots like me keep signing the Guardant form. I want private healthcare and physician freedom and novel, innovative therapies; the flip side of that is that physicians are financially incentivized to offer all manner of trash, whether it’s a trial or a legally marketed diagnostic; I want to hold patients (myself!) to the standard of not taking the bait.