Welcome back to Healthy Innovations! 👋

Across almost every major disease, timing decides more than treatment does. A diagnosis made months or years earlier can change everything that follows, and technology is finally catching up to that idea.

This week's Deep Dive follows a newborn whose genome was sequenced before her first symptom ever appeared, then widens out to AI reading eye scans for heart disease risk, a blood test replacing brain scans for Alzheimer's, and a multi-cancer blood test that just had a genuinely mixed first big trial.

Let's dive in!

When Margot was born in April 2023, her mother Lili enrolled her in a study most new parents have never heard of. Instead of the standard heel-prick test, Margot's genome was sequenced in her first days of life, screened against more than 250 early-onset and neurodevelopmental genetic conditions as part of the GUARDIAN study, based at NewYork-Presbyterian and Columbia University. Launched in September 2022 and now led by geneticist Wendy Chung from Boston Children's Hospital, GUARDIAN has since screened more than 15,000 newborns and expanded its panel to more than 450 treatable genetic conditions.

The sequencing flagged CDKL5 deficiency disorder, a rare condition that typically causes severe seizures and developmental delay. Margot had shown nothing.

Six weeks later, the seizures started exactly as predicted. But her care team was not starting from zero. They already knew what they were looking at, and had a monitoring and treatment plan in place before the first episode.

Margot's story used to be a research curiosity. It is quickly becoming a preview of where diagnosis is heading across several major disease categories, though not all at the same pace or with the same certainty.

For decades, healthcare has been built to respond: a symptom appears, a test confirms it, a diagnosis follows. What is changing, unevenly, is the order of operations. Over the last eighteen months, newborn genome sequencing, blood-based Alzheimer's tests, AI reading of routine scans, and multi-cancer blood tests have all pushed at the moment of discovery, moving it earlier for some patients and some conditions.

The harder question is not whether any of this works. It is how consistently it works, and whether an earlier signal reliably becomes a better outcome, rather than just an earlier answer.

Cardiovascular risk has traditionally relied on a handful of blunt instruments: blood pressure cuffs and cholesterol panels. Neither sees inside the blood vessels themselves.

Researchers at the University of Leeds and the British Heart Foundation have shown, in two stages:

A separate model developed at Israel's Technion Faculty of Biomedical Engineering, called DeepHHF, works from a routine ECG instead, identifying patients at high risk of heart failure years before clinical disease develops.

For decades, a definitive Alzheimer's diagnosis has meant a PET scan or a lumbar puncture: expensive, invasive, and often not offered until symptoms are already hard to ignore.

Two blood tests are changing that:

Neither test replaces a full diagnostic workup, but both mean a GP can now help rule Alzheimer's in or out using the same blood draw as a cholesterol check, rather than referring a patient into a specialist queue that can run for months.

Osteoporosis is usually caught the way disasters are: after the first fracture. Bone density scanning exists, but it is not part of routine screening for most adults, and it competes for time and budget in a stretched health system.

Research including work out of NYU Langone is exploring a different approach: mining CT scans that patients have already had for completely different reasons, such as a chest scan taken after a cancer diagnosis or an abdominal scan taken after an unrelated injury. Early results suggest AI models can extract vertebral and hip bone mineral density from these existing images with no new appointment and no additional radiation, flagging fragility risk years before a fall makes it obvious.

This is still exploratory work, not a deployed screening tool, but researchers estimate even partial rollout could catch a meaningful share of the millions of undiagnosed osteoporosis cases worldwide, simply by looking properly at data already being collected. For someone who will never book a bone density scan unless something breaks first, that diagnosis could arrive without them having to ask for it.

Cancer screening has always meant choosing which cancer to look for. Mammograms find breast cancer. Colonoscopies find colorectal cancer. Most cancers have no organized screening program at all.

Multi-cancer early detection blood tests, led by GRAIL's Galleri test, are built around a different premise: analyzing cell-free DNA fragments circulating in the blood for a signal shared across more than 50 cancer types from a single draw.

The most rigorous test of that premise reported mixed results in 2026. NHS-Galleri, the first randomized controlled trial of a multi-cancer early detection test, followed more than 142,000 people in England for three years. Findings presented at ASCO, showed it missed its primary goal of reducing the overall number of late-stage cancer diagnoses.

However, GRAIL's own reporting of the secondary results is more encouraging, though company-reported. Compared with standard screening, the trial found:

A separate US and Canadian study, PATHFINDER 2, found a similar pattern when Galleri was added to existing recommended screening: substantially more cancers detected, 71% of them caught at stage I to III rather than later. Reported multipliers for the increase vary from roughly six to seven times, so treat the exact figure as approximate.

Read together, the two trials suggest this generation of blood-based cancer screening is not yet shifting outcomes at a population level the way early supporters hoped. It is, however, reliably finding more cancer, and finding it earlier, than today's standard screening does. GRAIL submitted the final module of its FDA premarket approval application in January 2026, betting regulators will agree that is enough to build on.

Margot is now a toddler, developing under a monitoring plan her doctors built before her first seizure ever happened. Her story, and the technologies behind the other three, point the same way without yet adding up to one tidy narrative.

None of this replaces treatment, and the evidence is not uniform across conditions:

What holds across all four is timing. The moment of discovery is moving earlier, years before it would otherwise have found a voice, even while what happens after that moment still depends heavily on which disease it is.

👁️ Science Corp secured a CE mark this year for Prima, a wireless chip implanted beneath the retina paired with camera-equipped glasses, becoming the first treatment able to restore functional central vision in people with geographic atrophy, an advanced form of age-related macular degeneration and a leading cause of blindness in older adults. The glasses capture video and beam it to the implant as near-infrared light, which the chip converts into electrical signals the brain can interpret.

In trials, patients gained a mean of 25.5 letters on a standard eye chart, more than five lines, and 84% could read letters, numbers and words again. Some have gone further, recognizing faces, sketching from memory, and finishing full novels. The first commercial procedure is expected in Germany, where the trials were run, as early as September, with the device expected to reach the US as early as next year.

🧬 Redheads: Recessive not extinct. Less than 2% of people worldwide have red hair, which has fueled decades of rumors that redheads are on their way out. It's not a new worry: scientists were predicting the death of blond hair back in the 19th century too, and that never happened either.

Red hair survives because the gene behind it, MC1R, is recessive, so plenty of people with brown, black or blond hair are carrying a copy without ever knowing it. It takes two carriers to have a child before red hair actually shows up, which means the real number of people carrying the gene is much larger than the number you'd spot on the street.

The trait also isn't going anywhere for a simpler reason: for something to disappear from a population, the people carrying it need to have fewer children than everyone else, and there's no evidence that's happening. Even the higher skin cancer risk linked to red hair mostly shows up well after childbearing age, so it has little effect either way. There may even be an upside: the same pigment responsible for red hair seems to help cells manage a byproduct that becomes toxic at high concentrations.

Thank you for reading the Healthy Innovations newsletter!

Keep an eye out for next week’s issue, where I will highlight the healthcare innovations you need to know about.

Have a great week!

Alison ✨

P.S. If you enjoyed reading the Healthy Innovations newsletter, please subscribe so I know the content is valuable to you!