Back in the 90s during graduate school I was excited about an idea in cardiovascular imaging and talked to a stent lab cardiologist who had successfully started a number of companies. I told him my idea, and I will never forget his response.
“Steve, Wall Street treats these three mistakes exactly the same. Too early, too late, too stupid. This idea of yours is too late, sorry.”
Too early, too late, too stupid. 2E=2L=2S.
My first biotech I worked for was a Bay Area company in 2000 in the Al Zaffaroni lineage (Syntex-Roche, ALZA, Affymetrix, Maxygen), doing biomarker discovery in autoimmune disease with terabyte sized databases with patient clinical data, mass spec profiling (before it was called that), and high content cytometry. Very cool stuff, one employee even went out and purchased the domain www.biomarker.com. But ultimately, we were too early, by about 7 years. Pharma wasn’t ready for personalized medicine, it still barely is. The human population is barely ready for ‘personalized medicine’, and certainly has no clue how to pay for it.
While we were busy with our biomarker terabyte databases, a small company that also had 30 people showed up across our labs in the same office park in Mountain View. They were an internet search company, and at that time, if you asked 20 people if there was any money in internet searching, 19 people would have said no. Dead end, no one had made money at it. Freeway signboards on my daily commute were all were full of dead search engine companies — Magellan, Excite, Infoseek, Inktomi, Northern Light, and AltaVista.
We shared a common lunch area with this search engine start-up, and myself and another part computer scientist part biologist friend were asked if we wanted to work for them. “No way”, we both said, “You guys with your search engine will be out of business soon, and we are going to cure arthritis and multiple sclerosis (MS).” Our company grew to 70 people, theirs grew to 70 people, ours shrunk back to 30 people, theirs grew to 200 people. theirs grew to 400, 800…
The company was Google, we would have been employee number 34 and 35. Um, we will call that one Too Stupid.
Now back to digital pathology and technology adoption.
Despite all the fuss, neither the technology nor the pathologists are ready for primary read on digital slides. Not in GLP tox pathology assessment, not in H&E reading in a hospital. It will take 3 more years, not just to get the imaging right, but more importantly to get the pathologist workflow infrastucture right. Those cardboard sleeves and slide boxes, and LIS systems they use, those don’t just happen, it isn’t random glass and brass that will just get tossed out when digital slides show up. A good pathology practice has built the entire system around making the pathologist as fast, effective, and accurate as possible. Sit and watch pathologists, watch how quickly they read H&Es. Observe how the entire histology department provides them glass and paper at the right time and in the right way so pathologists most effective with their time. This is the workflow a good digital pathology vendor needs to emulate, and this is the next software that will sell well. Until then, forget about primary read.
But two areas where Flagship is spending its pathology resources — GLP image analysis and peer reviews. Neither require every slide to be read, both have economics and industry need on their side. Image analysis because quantitative efficacy and toxicity data is required for making huge go/no go decisions. Peer review because as the world gets smaller, travel gets more expensive.
Flagship has an exciting services alliance with EPL, the premier preclinical pathology company who can fairly be said to have pioneered peer reviews. We will make these peer reviews secure, cost effective, and fit into the pathology workflow. We have a group of 5 pharma and one medical device pathologist called VIPER, or Virtual Imaging in Peer Reviews, and a donated primate study that has been read and peer reviewed. We will look at organization, time, quality, scanning resolution (20x or 40x?), Z-stacking, and determine a validation plan for peer reviews.
Please contact us if your organization wants to participate.
With image analysis, we are working on GLP image analysis processes. It was almost three years ago that I gave a webinar proposing vendor requirements for GLP in whole slide imaging. The next logical step is GLP with image analysis. Not a trivial problem, but one that needs to be solved, and we will solve.
Too early, too late, too stupid. Let’s be on time! Let’s save time. Personalized medicine cannot afford the 10 year drug and device development timelines we have had in the past. The patients can’t afford it, the payers can’t afford it, the investors can’t afford it. Tissue assessment is one of the most difficult and time-consuming parts of drug and device development. We need to raise the digital standard.