Cancer Screens and Biotech Dreams
Reflections on diagnostic tests, mission-driven careers, and the fallacy of moral imperatives
I’m really excited to share today’s (anonymous!) essay, which examines what it really means to be working at a science- and mission- driven organization. Some of this essay is very relatable—who among us hasn’t gone through the angst of trying to find, and maintain, a ‘values-aligned’ job? In editing other parts of this essay, I learned a lot about an industry that I admittedly hadn’t thought very deeply about before. Biotech coverage at Reboot more broadly has been sparse; hopefully this piece is the beginning of more.
— Jessica Dai, Reboot Editorial Board
Cancer screens and biotech dreams
Anonymous
Several years ago, I was a bright-eyed computer science graduate attempting to find the balance between the impact-driven work I desired and the suitable application of my degree. Having spent years entrenched in Silicon Valley culture, I felt wary of selling out. But by the time I received my job offer, I was convinced that I would, in fact, be able to mature as an engineer while developing a product that was decidedly a positive impact: an early cancer detection test. Leadership had recently announced that the proprietary cancer detection technology—which involved searching through patterns at over a million genomic sites, from fragments of DNA shed into the blood from nascent and otherwise undetectable cancers—had received early regulatory clearances for use in a multi-cancer clinical screening test.
There had always been a stark homogeneity around some of the messaging. The office featured a brightly colored wall emblazoned with the company's slogan in a tacky font. At the same time, corporate transparency seemed apparent. Every new employee got a lab tour. We had frequent mock audits. And in what I’ll call the R&D times, before product launch, the science teams maintained an open culture around the early detection approach and the various technical and scientific issues we’d have to overcome, as a company, before the great potential of our product could be realized. With internal and external confidence about the underlying technology growing, the question was no longer “does the tech work?” but “is it providing value?”—no longer a question of can, but a question of should. After all, why would a physician decide to prescribe or trust such a test? What should be the recommended clinical workup for a positive result? What were the practical implications of the threshold for false positives chosen as tolerable? And what were the correct epidemiological metrics for determining the test’s effectiveness?
The scientific mission was a heady one, but everyone I’d ever spoken with was aware of the substantial practical constraints and the fact that early detection did not automatically translate into lives saved. Nevertheless, we giddily celebrated corporate milestones: one holiday party included the announcement that enrollment had begun for the first return-of-results study. This study was the first application of our technology that would actually send a report result to participants and their clinicians, enabling intervention—the ultimate goal of a screening test. This transition from observational studies, which were important for assessing test performance during development but didn’t provide actionable results to patients, to interventional ones was the first big shift I experienced as an employee.
I loved my engineering role, working in the office with a collaborative team, but we experienced a second big shift due to the global pandemic. After a few weeks—when it became clear that shutdowns would last months, and also that software teams were perfectly capable of carrying out their job functions from home—working from the kitchen table became my new normal. During this time, we gained new software leadership, marking a change in company priorities and a restructuring towards the launch of the flagship multi-cancer screening test. While R&D activities continued in the background, the corporate structure and quarterly goals centered around this new, commercial phase.
We were often reminded that, as employees of a mission-driven company, we were also its ambassadors—in hiring, at conferences, and to our friends, family, and acquaintances who may eventually take the screening test. Beyond the wall with the tacky lettering, there had always been a lingering message that we had a moral imperative to work hard for the company's success. This moral imperative was driven by a collective desire for impact, to make the screening product as widely available as soon as possible. Screening tests saved lives. But while this collective ethos originally felt organic, it later became corporatized: it was internal and external marketing to recruit us, to keep us focused on the mission, make us work harder, not complain when we had to work evenings and weekends. Like many other biotech companies (some much less evil than others), the mission happened to conveniently align with the capitalist constraint of maximizing worker productivity.
In the post-covid era, this somewhat inevitable shift—losing some of the nuance and transparency of the previously scientific work culture, gaining a commercially-oriented party line—became pointed as the company geared up for the first product launch. This shift meant that software teams were building systems for commercial applications rather than scientific ones. It means that the high volume of new hires across the company’s commercial divisions were simply less interested in the science than the employees who came before. And new employees were no longer given lab tours.
A screening test implies certain economics. By definition, a screening test is not for sick patients, but for use in a general population without any clinical symptoms. This is a key constraint in the business model of any company that makes screening tests. When there is money to be made on each screen, a profit-driven industry has an incentive to screen more people. This incentive is countered by the drop in how informative each test is if there are more false positive results.
This reality is quite literally a textbook example. Many students are introduced to Bayes' rule using a conditional probability example from medical diagnostics, specifically screening tests. When designing a test, we think about measuring performance by asking, “given that a patient has the disease, what is the probability that they will (correctly) test positive with our test?” But if I were a patient who tested positive, I might want to know, “given that I received a positive result with this test, what is the probability that I actually have the disease?” Writing out the mathematical formulation, it turns out that the prior probability of the screened population having the disease is crucial. Practically speaking, a company with a diagnostic product must carefully and narrowly define the population to be screened. Defining this prior for a diagnostic—for example, choosing to screen only those at a higher risk from age or environmental factors—is a crucial choice in lowering false positives.
So, does screening save lives? It depends on who you screen, and the history of how screening guidelines have changed in response to new diagnostics illustrates this point. When mammograms were first introduced, in the 1950s and 1960s, clinical trials reported mixed results on whether they prevented additional deaths. In the 1980s, the American Cancer Society recommended baseline mammograms for all women in their 30s, later raising the starting age for mammograms to 40 in 1992. Today, the American Cancer Society recommends that high-risk women get a breast MRI and mammogram annually starting at age 30. Various entities have modified their screening guidelines numerous times since the 1960s, chasing the best definition of a prior population on which to recommend screening, in light of updated evidence.
The general logic is: why screen earlier if the cancer is typically still curable, with life-saving intervention possible, later on? The proportion of false positive results, which can be extremely psychologically damaging (imagine rearranging your life after an incorrect diagnosis), is generally reduced by starting screenings later. Screening starting from a higher age therefore seems an effective way to narrow the screening population, as long as high-risk individuals are also screening starting earlier. Official screening recommendations continue to flip-flop, but at the very least, modern guidelines tend to emphasize incorporating personal risk factors.
A continued focus on early detection involves an assumption that we will be able to corroborate cancer-positive results and intervene early enough to make a difference. However, most cancer diagnostics can’t tell us anything about how aggressive a cancer is or how quickly it will grow. While the biology is still somewhat murky, we know there is high variability between individuals, and that growing cancers carry unique sets of mutations in different individuals. Many cancer treatments are very personalized. In my in-office conversations, my colleagues and I would chat about improved cancer therapies and how they could make early detection more valuable.
There are considerations for screening tests beyond clinical utility. One practical resource constraint is that, even if a person pays for screening out of pocket, the clinical workups would likely involve shared resources (e.g. CT machines). Is this an improper allocation of medical resources? More existentially, we could ask whether becoming reliant on frequent screening—looking for tiny, early signals of a cancer that may remain treatable even at later stages—is a good thing at all. Is it valid under the assumption that new therapies and earlier interventions will become possible as we learn more about cancer, and that these interventions will save lives that previous interventions couldn’t? Or is it simply a case of overscreening and overdiagnosing in the name of profiting the medical-industrial complex?
My own experiences made me wonder whether the value provided by cancer-focused companies is really in their huge cohort studies, from which we can learn enormous amounts of new cancer biology, including patterns and biomarkers of cancer progression. These large-scale studies are expensive; unlike academic entities, which are often limited to analyzing data from existing trials, companies can actually fund new studies, designed and run in partnership with hospital systems. While screening large populations is the focus, smaller scale and less profitable applications of cancer-detection technologies include their use as tools for tracking residual disease present in a known cancer patient after starting therapy. Data collected from such a cohort is extremely valuable for studying the dynamics of cancer progression and treatment—an important metric for many pharma companies, who develop therapeutics on this smaller scale.
Unfortunately, biotech companies can’t sway their investors with a roadmap of scientific discovery alone, even when it’s about cancer. Companies emphasize screening to have a viable business plan, meaning a product that will be broadly used by enough people to make a profit. Since biotech is mostly therapeutics or diagnostics, we should note that the former has a much higher cost to consumers. The math just works out better if healthy populations without any apparent symptoms are being screened with diagnostics, rather than only sick patients being treated with therapeutics.
Such is the hypocrisy of biotech. Many employees work hard in the name of scientific discovery and technological advancement, but projected profitability is a company’s bottom line, set by product roadmaps, stakeholders, and investors. My workplace fit the norm of a scientific company being pinned to its commercial interests in order to stay alive. But the moral imperatives invoked as marketing and motivation tactics, which rested on an axiom of early detection immediately translating into saved lives, were as hollow as the axiom itself. And yet, these moral imperatives were laid out publicly in press releases and product announcements, stated as self-evident, and frequently used to justify high-level corporate decisions.
I left the company several years ago. I don’t regret my experience or any of the company's achievements. I no longer have the uncritical view of “mission-driven” work or “impact” that I did as an undergrad, but I do still believe that a focused mission enables strong teamwork and provides enormous scientific value. Companies that develop multi-cancer screening tests provide completely novel hope for the cancers that don’t currently have traditional screening available, such as pancreatic cancer. Other cancers are currently screened for only in high-risk patients, such as lung cancer screening being done only for those with a smoking history.
If companies can do these screenings simultaneously and well, with nothing more invasive than a standard blood draw, it must mean something. If we can learn more about the underlying biology and dynamics of cancer progression and variability from an early, detectable stage, it must mean something. However, science is an inherently messy pursuit, at odds with cut-and-dry corporate measures of success. We should be honest that none of these somethings are a silver bullet for saving lives; to believe so is a fallacy, a moral oversimplification. But as long as they can develop the technologies, fund the trials, and collect the data that government or academia will not, biotech companies remain a bet worth taking.
Anonymous is a computer scientist with biotech and tech industry experience.
Reboot publishes essays on tech, humanity, and power every week. If you want to keep up with the community, subscribe below ⚡️
🌀 microdoses
Am I the only one getting nonstop “chat with your PDF” ads?
If you’re in the US, you have one more day to apply to work for the US AI Safety Institute1
I couldn’t pick just one Raygun tweet, so here’s a curated selection: Trader Joes, first-year grad students, and a long semi-serious discussion
Was there a ‘vibe shift’ in how tech thinks about politics? (Did it become socially acceptable to ‘come out’ pro-Trump?) Or was there not?
💝 closing note
To the last microdose: I am always looking for pitches on the vibes of a particular industry or subfield. You can be anonymous!
—Jessica & Reboot Team
we can talk about the ‘Safety’ terminology at another time….