2013 has shown healthcare is rapidly becoming an information science, a social science and a network science. One of my favorite stories of 2013 that showed this emergence was The Engagement Pill, based on Epilepsy community research by PatientsLikeMe, hinting that patient communities offered a dose-response for some patient measures. If we want to understand healthcare, how to deliver it, how to pay for it, and how to improve it, we need to understand networks, connections and how to grow them to work for us.
Another company leading the way, or was leading it before a letter from the FDA, was 23andMe. If you haven’t heard, on November 22, 2013, 23andMe received a letter from the FDA asking the company “to stop marketing its Saliva Collection Kit and Personal Genome Service (PGS).” 23andMe has genotyped, via SNP-analysis, more than a half million people through the test. If you need to catch up, there’s a great resource of links here to get up to speed at the end of this post.
23andMe offers a Personal Genome Service (PGS) and several communities for people interested in specific shared genotype markers, conditions or genealogy, but can no longer provide genomic analysis. 23andMe’s issues with the FDA stand out as the most fascinating and complicated stories of 2013 for its potential impact on healthcare information science, citizen science, social science and network science. As we begin 2014, we need to think hard about how to proceed toward a networked health future.
As an active proponent of people taking control of their health with information, my first reaction upon hearing about the letter was: How could the FDA separate people from their own genetic info? I’ve now dug a little deeper as the dust has settled and I now see it as a complicated story and an unfortunate series of events with a lot of missed opportunities and plenty of conflicting interests.
There’s a cornucopia of questions related to the data-driven, social network-linked and networked future of medicine that 23andMe was actively creating. There are questions about the business model; questions about the regulatory authority of the FDA; and questions on developing a healthcare community that shares data, touching legal, ethical, privacy, medical, and statistical issues just for a start. Add to all that a soap opera narrative of broken relationships and intrigue and you’ve got a saga worthy of a miniseries.
I believe mistakes were made on both sides, but I see a company that was trying to be too many things for too many people, serving many conflicting masters.
Putting on the brakes as we seek to answer some of these questions may not be such a bad idea. This is a big moment in the history of medicine as we try to come to terms with a social, big data era.
“Where do we come from? What are we? Where are we going?”
There exists a minefield of powerful forces of big data, identity, privacy and security. There will be a delicate dance to navigate them for societal benefit. Walking the middle line between group coordination and individual benefit is never easy. It’s been a part of our human history from the beginning, as E. O. Wilson writes in “The Social Conquest of Earth.” Our own human species has been unique in this capacity to manage group and individual benefit in unique ways – it’s in our DNA (the same DNA sequences that are at the heart of this story, in fact), and we’re now in the process of baking these systems into our societal DNA, and, just as Wilson quotes Guagin in the beginning of his book, we need to ask where we are going.
In order to succeed, much like our ancestors, we’ll need to do it in a way that benefits and offers control to the individuals who participate. Only now can we share data resources across the world in an instant. We’ll need to simultaneously serve the interests of those who provide the data, those who analyze it, and those who create value from it. Sometimes those may be the same people, often different. We must find a way to share information that protects privacy and leads to greater discovery. There’s just too much value and too many lives to be saved not to get there.
So let’s look at some of these issues in three parts. First, problems surrounding 23andMe as a company and a business model that might hint at how to do things better, and I’ll focus on issues I haven’t seen explored in detail elsewhere. In part two I’ll discuss issues within the regulatory framework the company is attempting to work (and not built on a social model). In the final post I’ll suggested a path forward and highlight some of the issues that will need to resolved.
Part 1. A Conflicted Business Model in Personal Genomics
The FDA has taken issue with the claims 23andMe made on the genetic analysis, saying that “23andMe had not demonstrated that they have ‘analytically or clinically validated the PGS for its intended uses’ and the ‘FDA is concerned about the public health consequences of inaccurate results from the PGS device.’”
While the FDA points to not demonstrating validation, a greater issue may have been an overall lack of transparency. After all, competitors offer differing versions of the analysis, including SNPedia and sister software promethease.com, which offer a test for $5 to which you can submit the SNP data you can still get from 23andMe. These services haven’t received letters from the FDA.
The analysis you can get on Promethease doesn’t appear to be all that compelling, either. Still, it appears that as an open wiki, those problems can be solved by the crowd to improve the product, so there is a great deal of transparency in the analysis. And with only a small charge for the service, there may be less concern to the FDA about overselling the analysis and benefits — certainly less of a perceived need to hype the benefits.
Perhaps charging for the analysis was the wrong move, more about that in a minute, but there is also a more subtle lack of transparency. I believe it hints at a conflicted business model.
According to a Fast Company article before the FDA letter:
“23andMe’s privacy statement clearly states that it collects a person’s genetic, registration, web browsing, and self-reported information. The company can share its data with third parties [after] it has been stripped of Registration Information and combined with data from a number of other users sufficient to minimize the possibility of exposing individual-level information while still providing scientific evidence.”
This statement leaves the door open for 23andMe to link genetics to online and other behavior. As the article points out, minimizing is no guarantee. I suspect, even if it’s stated clearly in the privacy statement, this isn’t very well-informed consent for most users of the service. How much do each of us know about how Facebook or Gmail use our data?
Building and Selling the Data
The long-range plans weren’t a secret. It’s about a big data set, not selling the kits and analysis.
From a Fast Company article written in October, before the FDA letter:
“The long game here is not to make money selling kits, although the kits are essential to get the base level data,” says Patrick Chung, a 23andMe board member and partner at the venture-capital firm NEA. “Once you have the data, [the company] does actually become the Google of personalized health care.”
23andMe is halfway to their stated goal of genotyping more than a million people, yet some signs indicate they were struggling to sell the value of their intended business model. According to a pharma exec in the same article above: “No one has figured out the right dollar amount and the right revenue model to pay for that online community’s rich source of data.”
Perhaps Pharma knows that the predictive powers and therefore the potential value of SNP analysis was limited, as many customers have attested.
Was this kind of response inhibiting them from committing to a free model? They likely weren’t covering their costs on the free model, so why even bother charging? Would 23andMe have suffered the same fate if it had given the test away for free?
Charging for the Kits
23andMe has raised $126 million. That, in essence, means investors believe there’s at least a $1 billion or more worth of potential (most early investors are looking for a 10x return or more), and they (admittedly) know they won’t get there by selling $99 tests. They’ll get it from mining all the data (SNP as well as other kinds of data in the future) from customers across the world. In 2012, after raising an additional $60 million, they lowered the price of their “spit-test” analysis from $299 to $99, after experimenting with selling them free of charge.
It’s been said that “if you don’t pay for the product (like Facebook or Gmail), the product is you.” In 23andMe’s case, people still paid for the product while customers remained the main product. It seems rather bold to charge for a service involving personal, genetic data and then use that purchased data to build yet another asset. Google doesn’t charge people to have their sites crawled for its search engine. If you want to build this kind of asset, don’t charge for it.
The closest model for this kind of business, perhaps, would be PatientsLikeMe. However, PLM 1) doesn’t make claims, 2) doesn’t charge, 3) doesn’t collect genetic info, 4) sells data in a very up-front manner, and 5) information is provided voluntarily by customers.
Who benefits? Walking the line between citizen science for community benefit and citizen science for company benefit
“..the one characteristic that set it apart from other services offering genetic testing beyond the clinic is that they have tried to include ‘citizen science’ elements and encourage wide data sharing. This has played a relatively marginal role in the wider discussion. We are not uncritically applauding the way that they have been doing this, but it adds an important new dimension and direction for science.”
(There’s another half to this post that I’ll include in my next post on the FDA.)
I applaud the use of citizen science, but let’s make sure citizen science is done for the citizens and that the model is one of more open platforms. We need informed consent, and we need to ensure that citizen science means that the citizens can share in the results, that it’s open in a privacy-aware way.
23andMe has sold “becoming a part of our community,” yet you have to pay a monthly subscription to remain a member of the community.
You always have to be suspect, as Brad Feld has pointed out in “Startup Communities: Building an Entrepreneurial Ecosystem in Your City” of people who refer to something as “our community.” The community, if it really is one, belongs to, and is led by, the the community members, nobody owns it. But that was never part of the deal. When you pay to donate your data to 23andMe, it, and the benefits they impart, are up to them to share, and you keep getting receiving so long as you keep paying.
It’s easy to suspect this wasn’t all for finding the best treatments.
Support the community, please, but don’t call it a community when it’s a club. I’m all for competition between walled gardens, completely proprietary systems and open platforms, but let’s not sell one when we’re actually signing up for another.
Jumping the Gun
“In one study that would’ve taken the medical establishment tens of millions of dollars and up to a decade of research, 23andMe was able to analyze all that data and identify two novel genes that are highly correlative to people who have Parkinson’s. But the study also identified this odd group of people who were predisposed to Parkinson’s but were not symptomatic. So the research team created another community of customers and identified a gene that might possibly be protective.”
That’s great, assuming it’s more than just correlative, but it also highlights a key problem with the company: jumping ahead of itself. How much evidence was there that the other gene might possibly be protective vs a statistical anomaly? Or a correlation without causation? People will sign up, participate and be excited to take part in such a community, but if it’s not real, or the correlation turn up something embarrassing, there is a lot of false hope generated compared to a research study, where we know the researchers are seeking answers, not providing them.
I appreciate the desire to find a quick answer, but when is a community appropriate vs a research study? These kinds of quick steps to build a community look like jumping the gun and are suspect use of the term “community,” particularly when one of the early investors has an increased risk for Parkinson’s. These quick draw actions are the sort of thing the FDA is tasked to protect us against. I don’t know if people had to pay to be a part of this particular community, but that would certainly raise other issues.
Finally, what actions can people really take?
If I have a 20% increase of cardiovascular disease, can I do anything about it beyond what I should be doing already? Without further study, do we know which individual actions will influence their predetermined genetic risks, or are we just sending them on a goose chase of dread that they’re not doing enough?
As one doc has said, “Want to predict your health future, step on a scale.”
23andMe touts that you can take control, but can you? It’s presumptive to think you can really do a lot about your genetics other than direct new research now that you have a vested interest, exactly what 23andMe has seems to have done with their Parkinson’s community.
The FDA is also in uncharted territory, some argue about whether they belong in these waters. What kinds of questions does this raise for PGS tests going forward and home testing in general? That’s the topic of the next post.
We all have a stake in this. Our health data is a national (and international) asset, and we all have a shared interest in making it valuable as a collective. One of the great struggles of our time is how to make data bigger (and more valuable) by facilitating the sharing while protecting the interests of individuals who share it. I hope you’ll comment and make this a discussion on this series. We’re at a critical juncture, and we need to get this right.
For a great collection of links and stories on the 23andMe story check out David Dobbs link page.
Latest posts by Leonard Kish (see all)
- It’s Not Just a Watch, the World is the Interface in the New Behavioral Economy - October 1, 2014
- From Carepass to HealthKit: Health Platforms Must be Founded on Trust - September 10, 2014
- The High Cost of the Sweet Life - August 26, 2014