“Certain moves made long ago had produced all of them.”

— Kiran Desai, The Inheritance of Loss

Medicine has become obsessed with data in most good ways. Risk scores decide medication regimens. A biological age clock reads your methylation and rate of aging. An algorithm decides who gets flagged for screening and who doesn’t. I’ve built calculators like these myself. But more data is never the same as the right data, and even the right data is only carries half the story. The other half, what concerns me more, is the outcome that data connects to. does this number change what happens to the person in front of you? That question, what actually works for South Asians, and what doesn’t, is the whole reason Zinda exists.

Watch what happens when the right data isn’t there.

The UK’s standard cardiovascular calculator, QRISK, adjusts your risk score by whether you are Indian, Pakistani, or Bangladeshi, the same adjustment whether you arrived at 32 or were born down the road. The American pooled cohort equations were built on White and Black patients; the rest of us are mapped to the nearest approximation, in a silent gray invisble to metrics. Location, duration, the timing of every move: none of it is an input. So the coefficient knows the label. It cannot see time. Temporality is lost.

If our metrics are going to run medicine, and they are, they will have to learn to carry temporality: where a person lived, for how long, and at what age each environment arrived. This essay is about the system where that blind spot is most visible, and most fixable. The immune system keeps a legible record of place and time, written in markers we already know how to measure and outcomes we already know how to count. The heart and our metabolic machinery almost certainly keep similar records; the immune system is simply where the record is easiest to read, and where the feedback loop is easily established..

Picture a family at dinner in Los Angeles. The grandmother grew up in Delhi and came to California at 32. Her daughter made the same trip when she was seven. The grandson was born in LA (unfortunately a Dodgers fan) and knows Delhi through summer trips and voice notes. They eat the same dal and tell the same stories. In the database, all three fit into one categorical label as South Asian.

That word isn’t wrong. It carries ancestry, kinship, a family history that genuinely runs through three bodies. What it can’t carry is time. The grandmother’s immune system spent three decades learning Delhi before it ever met California. Her daughter’s was mid-education when the environment changed under her. The grandson’s has only ever known this one.

An immune system keeps a longer record than a passport.

For the first time, we have the tools to read that record: perform longitudinal biomarkers, intakes that can actually ask, models that can synthesize and hold a timeline. What was before a blurred immune canvas, gets painted with fine nuance and granularity. Telling a story of a person in ways we have never been able to see.

This series is about what they should be looking for.

One of the clearest clues comes from Ontario, where researchers followed 443,265 South Asian immigrants alongside other immigrants and non-immigrants. Immigrant children had less asthma than non-immigrant Canadian kids (incidence rate ratio 0.66, 95% CI 0.62 to 0.71) and less inflammatory bowel disease (IRR 0.47, 0.33 to 0.67).

Then you turn the page to their Ontario-born children, and the picture flips. Asthma climbs above the Canadian rate, IRR 1.75. The IBD advantage thins to nothing, point estimate 0.90 with a confidence interval crossing the line of no effect.

Same ancestry. Different answer, in one generation.

A companion analysis of over two million immigrants from everywhere found each decade younger at arrival came with roughly fourteen percent higher IBD risk. Observational, not causal, and not SA specifically which is almost the point. It looks less like a South Asian finding and more like a migration finding we happen to be well positioned to study.

And the pattern isn’t only Canadian. In 4.5 million UK primary care records, South Asians were diagnosed with immune-mediated disease about six years younger than white patients. Diagnosed, not struck and records like these can’t separate the two.

The Crohn’s data is quite intriguing. The second generation was diagnosed youngest, with disease that looked more like the Western inflammatory pattern.

• Native South Asians more often presented with the stricturing, complicating form (B2): the version that narrows the bowel.

• The second generation was diagnosed youngest (median ~17.7 years) and was significantly more likely to show the non-stricturing B1 inflammatory pattern (OR 3.48), i.e., their disease looked like what gastroenterologists typically see in Western clinics.

Whatever migration does to the SA phenotype, it doesn’t make the disease fiercer.

It makes it different.

Which exposure did any of this? I wish it was that clear cut, but with the data we have its impossible to untangle antibiotics from air quality from what was in the microbiome, and I’d rather sit in that unknown than reach for a tidy genetic story.

Proportion, not power

Immune vocabulary is derived from war.

Defenses, invaders, surveillance, cells recruited to a front. Fine during an infection. Useless the moment the real question becomes judgment. And likely why so much is missing in this understanding.

Inflammation closes wounds and keeps you alive; the same machinery injures you when it runs too long or fires in the wrong tissue, think autoimmune disease or inflammatory arthritis. What you want from an immune system isn’t strength but proportion:

• What belongs?

• What threatens?

• How much force?

• When to stop?

Immunologists call it calibration or learned judgment, built encounter by encounter out of everything the system has ever met. Each exposure leaving its footprint behind.

Migration doesn’t wipe that education. It adds a second one to further calibrate our immune systems. Edward Said described exile as a contrapuntal life, old place and new place sounding at the same time.Both lines keep playing.

So the history (and thus data) I want has an order to it:

1. What you inherited: genetics

2. What taught your immune system early: locations

3. What changed around you, and at what age: AQI, exposures, travel

4. What your tissues are reporting right now: biomarkers

5. Whether whatever got switched on can switch back off: treatment response and malleability

Seven is not thirty-two

People skip this part.

At thirty-two the grandmother’s immune system had done most of its learning. At seven her daughter’s was still in the middle of it. You can watch the difference in the gut: a Canadian study found longer residence and earlier arrival tracked with stool metagenomes closer to the Canadian-born group, down into function. Cross sectional data, but useful as the gradient is there.

It shows up in allergy too: among Kerala physician families who moved to North America, allergy rose in the same adults after migration, then again in their North America-born kids. One caveat is that this was self-reported data, in one professional community. But rhymes with the Ontario data above.

DNA travels between generations. So does immune education, through kitchens, households, neighborhoods, and the age at which each arrives.

That’s why, as we previously discussed in The Migrant Microbiome, migration is a biological event inside the gut, not a line in the social history.

The route is the variable

“South Asia to the West” describes almost nothing.

• An engineer going Bengaluru to San Jose

• A construction worker going Nepal to Qatar

• Families landing in Bradford, in Melbourne

• An Indo-Caribbean family in Queens

These are not one exposure. They are multifactorial, compounding across places, seasons, and hours of the day. And more than any clinician can hold in their head at once. Holding that complexity is exactly where machines esp AI could help. We’ll dive deep into that in parts 3 and 4.

Take sunlight.

Sylhet sits at 25 degrees north; London at 51. In Sylhet, the UV index stays high enough for skin to make vitamin D in every month of the year. In London it sinks below that threshold around October and stays there until spring, a vitamin D winter that no amount of lunchtime walks can fix. Skin built for the first place, now living and trying to adapt in the second: when studied on this, first-generation British-Bangladeshi women in London carried vitamin D levels of about 32 nmol/L vs 51 for the women who stayed in Sylhet, and 55 for their white British neighbors.Supplementation is vital.

Acculturation refuses to be a dial that turns one way. In MASALA, the biggest group, 73.8% of 771 participants, carried the lowest cardiometabolic burden while staying integrated into both cultures; the higher-risk classes were the more assimilated ones.

While UK Biobank flips this story: the highest risk sat with the most recently arrived, under five years in the country, a hazard ratio of 3.44, against 2.22 for the British-born and 2.14 for the longer-settled. Though I would argue this was a much smaller subgroup, about 400 people, and a metabolic outcome

Mostly it convinced me of this:

‘When’ someone arrived matters in some pathologies while in others may be a protection. Assimilation of diet, higher saturated fats and sugars might harm, but incorporation of athletics and a more active lifestyle might carry benefits not previously attained such as muscle mass and mitochondrial health.

What fat tells the immune system

The immune system listens to viruses and bacteria. It also listens to us:

• Stretched fat cells

• A leaky gut barrier

• A liver storing things it shouldn’t

Here the metabolic story and the immune story stop being separable.

Back in 2001, a West London study of 113 healthy adults found higher CRP in South Asian women, and, the finding that lasted, CRP tracked with visceral fat even after adjusting for BMI, while its links to insulin and lipids dissolved once visceral fat entered the model. A 2025 cohort of 322 Asian Indian adults in northwest India found more liver fat, larger visceral adipocytes, and a higher inflammatory macrophage ratio in those with diabetes, albeit small cohort nestled in a specific region.

Neither study says weight is irrelevant. But both point to an underlying implication of adipocytes in the immune process, weight matters, and why targeting lower BMI matters.

The part we don’t know

Immunology used to draw a clean line: T cells and B cells remember, everything else just reacts. Trained immunity smudged that line, and this is the chapter where it’s easiest to lie to ourselves.

The famous experiment fed mice a Western-style diet and found lasting changes in their myeloid precursors through the NLRP3 inflammasome, with supporting work in human monocytes ex-vivo. Mice, though were Ldlr-deficient ones. It doesn’t show that a diet permanently rewrites a person, and it is absolutely not evidence for a South Asian risk gene, because we don’t have one of those.

The human evidence is stranger and smaller when it comes to immune memory. Five volunteers were deliberately infected with malaria, then cured; their monocytes went quiet during the infection, then came back primed, with epigenetic marks to match.

Even there the memory faded by seven months with epigenetic markers wearing off.

The clinical question underneath stays good anyway:

What has this immune system learned, what is it hearing now, and can it stand down?

What the model should ask

Read more

For a South Asian, the answer is almost certainly no.

Roughly 8.8 million Indians already have dementia, and the number is projected to climb steeply as the population ages in the LASI-DAD national estimate.

Globally, dementia cases are forecast to nearly triple, from 57 million in 2019 to roughly 153 million by 2050, with the smallest increases in high-income countries and the largest in the low- and middle-income regions that include most of South Asia in the GBD 2050 forecast. The same trajectory is already visible across the region, where dementia prevalence among older adults runs high, around 8% in Bangladesh, for example, in a national older-adult dementia study.

Yet APOE ε4, the most visible genetic risk marker in Alzheimer’s conversations, is present in only about 10% of South Asians. That is roughly half the European frequency (~15.5%) and a third of the African frequency (~26.8%) in a LASI-DAD APOE frequency analysis.

So if the gene most associated with Alzheimer’s is twice as rare in South Asians, why is dementia still rising so fast in our population?

It changes what we should screen for first.

The APOE ε4 Paradox

APOE ε4 is the strongest known genetic risk factor for late-onset Alzheimer’s in European populations. If you carry one copy, your risk roughly triples. Carry two, and you are looking at a ~12-fold increased risk in a large APOE risk synthesis.

The ε4 variant is the centrepiece of many genetic conversations about Alzheimer’s risk. Commercial reports and clinical panels built around European-derived data often make APOE the most visible genetic signal for Alzheimer’s prediction.

India’s largest dementia cohort, LASI-DAD, makes the paradox harder to ignore:

1. APOE ε4 frequency in Indians is ~10.8% in LASI-DAD. That means ~90% of South Asians would not be flagged by the most visible APOE ε4 signal on European-calibrated panels.

2. When ε4 IS present, it hits hard: in Indian cohorts, ε4 carriers have been reported to carry roughly 5.9 times higher odds of dementia (OR 5.90, 95% CI: 3.44–10.13) - a figure highlighted in the LASI-DAD deep-genomics analysis. The lever works. It is just pulled less often.

3. European-derived genetic risk scores barely work here: A study testing 56 known AD risk SNPs discovered through European GWAS found that genetic risk scores explained only 0.1–0.6% of variance in Indian episodic memory in a South Asian polygenic-risk replication study. The authors concluded that “genetic factors found predominantly through EA-GWAS may play a limited role in South Asians.”

4. South Asians have their own risk genes: The same 2026 LASI-DAD deep-genomics study screened 3.43 million rare and 100,618 common “Indian-enriched” genetic variants, variants more common in India than in European, East Asian, or African populations, for association with cognitive function. Implicated genes like ASAH1-AS1, DENND4A, and RAB11 do not appear on any European risk panel. Sex-specific effects were dramatic: one male-specific variant (PHF21A) was associated with a 10.66-point drop in HMSE score.

In plain language, South Asian Alzheimer’s risk does not map cleanly onto the European model.

If you screen a South Asian with a European-derived APOE panel and get a “normal” result, the test may still be missing the biology that matters most.

This is not just an amyloid story

This is where the South Asian pattern breaks from the Western Alzheimer’s playbook.

In European and North American clinical teaching, Alzheimer’s disease is amyloid-dominant. Vascular dementia is usually treated as the second major category, often estimated around 15–20% of clinically diagnosed dementia, with AD responsible for 60–70% (vascular dementia clinical review).

In South Asia, vascular cognitive impairment accounts for roughly 40% of the dementia burden according to an India-focused vascular cognitive impairment review. That is roughly double the Western estimate.

A northern Indian tertiary centre’s study of 1,876 dementia patients found the AD-to-vascular-dementia ratio was approximately 1.1:1 — nearly equal. When you add the 21% classified as “mixed dementia,” the combined vascular contribution approaches 50%.

The EDIS study from Singapore, comparing ethnic Indians, Malays, and Chinese, found that older Indians had a higher prevalence of cognitive impairment (~24.6%) than Chinese, and remained more likely to be cognitively impaired even after adjusting for cardiovascular risk factors in the population-based EDIS analysis. A companion EDIS neuroimaging study found Indians carried a higher burden of subcortical atrophy and distinct cerebrovascular and neurodegenerative imaging markers than Chinese in a Scientific Reports imaging paper.

And the BRAINS study - a landmark comparison of 1,126 Indian South Asians, 1,176 British South Asians, and 1,736 White British stroke patients - confirmed that South Asian stroke onset occurs 7 to 19 years earlier than in White British populations in the BRAINS stroke cohort comparison. And across South Asian stroke cohorts, the phenotype skews toward small-vessel disease rather than large-artery atherosclerosis.

The pattern is hard to miss: in many South Asian patients, brain aging is arriving through a microvascular pathway first. The conventional Alzheimer’s screen, built around amyloid plaques and APOE ε4, is tuned for a disease that often starts somewhere else.

Two Engines of Cognitive Decline in South Asians

With the data in front of us, I want to lay out what I think is the most clinically useful framing. South Asian cognitive decline is powered by two converging engines. Neither of them depends on APOE ε4.

Engine One: The Metabolic Sink — “Type 3 Diabetes” And The IDE Competition

In 2005, Suzanne de la Monte and colleagues at Brown University published something that changed how neurologists think about Alzheimer’s in the original Type 3 Diabetes paper.

They showed that Alzheimer’s brains have extensive abnormalities in insulin, IGF-1, and IGF-2 signalling. Gene expression of these critical metabolic peptides and their receptors was reduced. The PI3K/Akt pathway, the backbone of insulin signalling, was impaired. The downstream consequences (increased GSK-3β activity, tau hyperphosphorylation, amyloid accumulation) looked like a metabolic disease, not a neurodegenerative one.

They called it Type 3 Diabetes: a brain-specific form of diabetes where insulin resistance and deficiency develop primarily inside the brain, overlapping with or occurring independently of the body-wide insulin resistance you measure with HbA1c.

This matters because it gives us a cleaner way to think about cognitive risk in South Asians.

One common South Asian branch is the insulin-resistant, compensatory-hyperinsulinaemic phenotype: lower BMI thresholds for metabolic syndrome, higher visceral adiposity at a given weight, sharper post-meal glucose spikes, and lower skeletal muscle reserves. But that is not the whole story.

The Zinda framework also assumes a Fragile Engine: beta-cells with lower reserve and earlier failure under load. Some patients therefore spend years in compensatory insulin excess, while others tip earlier into insulin-deficient phenotypes such as SIDD or Type 5 diabetes.

That distinction matters because the IDE mechanism applies most cleanly to the first branch: the insulin-degrading enzyme competition model.

Insulin-degrading enzyme (IDE) is a protease that your neurons rely on to clear amyloid-beta from the brain. IDE degrades both insulin and Aβ, using the same catalytic chamber. Crystal structures published in Nature in 2006 showed insulin and Aβ bound within the same IDE substrate-recognition system in the IDE crystal-structure study.

In a compensatory hyperinsulinaemic state, insulin can dominate that workload. The enzyme cannot distinguish between its two substrates. Insulin has an affinity for IDE (Km ~85 nM) that is roughly 300-fold higher than Aβ (Km ~25 μM). When insulin exposure stays high, IDE may spend more of its time clearing insulin, and Aβ clearance may suffer.

This is not theoretical. Farris and Tanzi showed in 2003 that knocking out the IDE gene in mice produced a >50% decrease in cerebral Aβ degradation, alongside hyperinsulinaemia and glucose intolerance in an IDE knockout study. Watson and colleagues then demonstrated in humans that experimental insulin infusion under euglycaemic conditions increased CSF Aβ42 levels, evidence that acute insulin exposure can modulate amyloid biology in vivo in a human insulin-infusion experiment.

The practical implication is narrower, but stronger: for South Asian patients who spend years in an insulin-resistant, compensatory-hyperinsulinaemic state, amyloid clearance may be impaired through the IDE bottleneck long before overt diabetes is diagnosed, regardless of APOE genotype.

Sarcopenia belongs in this picture too. South Asians carry some of the lowest skeletal muscle mass of any ethnic group. Muscle is the primary peripheral glucose sink. Mendelian randomization analyses support a causal relationship between low muscle mass / sarcopenia and cognitive impairment, with appendicular lean mass and walking pace causally linked to cognitive function in bidirectional MR. A Korean amyloid PET study found that higher thigh muscle mass was associated with significantly lesser risk of amyloid positivity (OR = 0.27 in females) in an amyloid PET body-composition study.

Engine Two: The Microvascular Burn And Crash

This is where Zinda’s framework maps directly to brain health.

We talk about the Burn and Crash across cardiovascular and metabolic domains. The same engine drives cerebral small vessel disease.

The SABRE study, which compared South Asian, European, and African Caribbean populations in the UK, found that diabetic South Asians had 63.3% more white matter hyperintensities (WMH) than diabetic Europeans in the SABRE brain MRI analysis. In the deep white matter regions, the excess was 102%. Each year of age led to an excess of 3.8% WMH load in frontal regions in South Asians compared to Europeans. And diastolic blood pressure predicted WMH in South Asians but not in Europeans — which means the vascular sensitivity of the South Asian brain to blood pressure is fundamentally different.

White matter hyperintensities are not incidental findings on an MRI.

Far too often we physicians treat these are chronic changes with no intervention to consider. That is a lazy practice of medicine. Reactive, not proactive.

These hyperintensities are evidence of chronic ischaemic damage to the brain’s wiring, the communication cables that connect one region to another. When your white matter is accumulating lesion load, your brain is losing its ability to route information efficiently. Processing speed drops. Executive function erodes. Memory retrieval slows. Eventually, the accumulated damage crosses a threshold where clinical dementia manifests, even in the absence of heavy amyloid burden.

Cerebrovascular reactivity, the ability of brain blood vessels to dilate in response to metabolic demand (measured via CO2 challenge testing), is emerging as an independent predictor of cognitive decline. A study in Neurology found that impaired CVR predicted cognitive performance independently of CSF Aβ42 and tau levels in a CVR biomarker study. CVR was a stronger predictor than white matter hyperintensity burden or vascular risk scores. A prospective study of adults with hypertension found that higher CVR in Alzheimer’s signature regions (precuneus, posterior cingulate, hippocampus) was associated with 40% lower risk of MCI in a hypertension CVR cohort.

A lipidology nuance: lowering ApoB does not starve the brain

There is a common fear in healthspan/preventative medicine circles: “If I lower LDL too much, am I depriving my brain of cholesterol?”

This is the wrong model.

The brain is cholesterol-rich, but it largely runs its own cholesterol economy. Peripheral apoB-containing particles like LDL do not meaningfully deliver cholesterol into the brain because the blood-brain barrier separates plasma lipid traffic from central nervous system lipid traffic. Brain cholesterol is synthesized locally and trafficked through apoE-containing, HDL-like particles made largely by astrocytes and other glial cells. Adult neurons then acquire cholesterol through apoE-recognizing receptors such as LDLR and LRP1 (brain cholesterol homeostasis review; brain lipoprotein review).

Peter Attia went deep on this topic with Tom Dayspring this past week (the main reason this post was delayed).

That changes how to think about APOE4. APOE4 is not simply a “bad Alzheimer’s gene.” It is a different apoE protein shape in the brain’s lipid-transport system. If apoE4-containing brain lipoproteins traffic cholesterol less effectively, neuronal membrane cholesterol homeostasis can become unstable. And membrane cholesterol matters: experimental work shows that astrocyte-derived cholesterol can regulate neuronal amyloid-beta production by changing how APP interacts with secretases (PNAS cholesterol-amyloid study).

So the practical conclusion is almost the opposite of the folk fear. For a high-risk South Asian

patient, lowering ApoB and Lp(a)-linked vascular risk is brain-protective through the microvascular pathway, not brain-depleting. The safety literature is reassuring: a 2025 systematic review and meta-analysis of lipid-lowering trials found no significant association between lipid-lowering therapy and dementia or cognitive impairment, while confirming the need for better dementia-specific trials (lipid-lowering cognition meta-analysis).

The nuance is not “never worry about brain cholesterol.” It is: do not confuse plasma LDL-C with brain cholesterol supply. In select high-risk APOE4 patients, advanced clinicians may track emerging sterol markers such as desmosterol or 24S-hydroxycholesterol, but those are refinement tools — not a reason to leave ApoB high.

For the source conversation behind this framing, see Peter Attia’s episode with Tom Dayspring:

The screening stack: what Zinda would measure earlier

The conventional paradigm asks you to wait until you have memory complaints, then run a cognitive assessment and, when amyloid-targeting therapy or specialist risk stratification is on the table, discuss APOE ε4. For South Asians, that is too late. By the time Alzheimer’s disease manifests clinically, the underlying pathology has been accumulating for 15 to 20 years.

Read more

You book the scan, lie still for a few minutes, and a number comes back: 0.

For many people, that result is genuinely reassuring.

But for South Asians, the deeper question is whether a zero calcium score sees the whole coronary-risk story, especially when family history, symptoms, diabetes, ApoB, Lp(a), or insulin resistance are already raising the signal.

We now have technology to provide nuance in a way never previously available.

This is the kind of question Zinda Futures is built for.

Medicine is moving from simple risk markers toward richer imaging, AI-assisted plaque quantification, and more mechanism-aware prevention. That could be a major advance, especilaly if the tools are calibrated to the biology of the people using them.

For South Asians, calcium scoring can be useful yet still incomplete. A single reassuring number can become the end of the conversation while the biology is still moving underneath it.

A CAC score of zero is one of the strongest negative risk markers we have in preventive cardiology. It can help avoid unnecessary medication in the right person. It can calm down a borderline risk conversation. It is quick, relatively inexpensive, and widely available.

For South Asians, that result should sometimes lead to a second question, because CAC answers a narrower question than many people realize.

A calcium score asks how much plaque has hardened. It is less direct at answering how much plaque is actually there.

And that distinction may matter a lot for us.

What CAC Actually Measures

Coronary artery calcium scoring measures calcified plaque in the coronary arteries. It is looking for the hardened, mineralized footprint of atherosclerosis. When that footprint is large, risk rises. When it is absent, risk is usually lower.

That is why CAC has become so useful in primary prevention. In someone with borderline or intermediate risk, a CAC score can help decide whether to start or intensify preventive therapy. The 2019 ACC/AHA prevention guideline explicitly lists South Asian ancestry as a risk-enhancing factor and supports CAC testing when risk remains uncertain after the usual calculation.

The same guideline also makes the important point that the absence of CAC does not rule out non-calcified plaque.

That sentence is easy to skip, but it carries the whole argument. Non-calcified plaque is atherosclerosis before it has calcified. It can be lipid-rich, inflamed, narrowing the artery, and still invisible to a calcium score.

A zero calcium score means the scan did not see calcified plaque. The artery wall may still deserve attention when the clinical story is concerning.

For many low-risk, asymptomatic patients, that nuance may not change much. CAC=0 is still a good sign.

In a South Asian patient with chest pain or premature family history, diabetes or insulin resistance, high ApoB, high Lp(a), central adiposity, or a story that just does not fit the number, that nuance becomes the whole point.

We have seen this pattern before in other parts of the Zinda framework. In The Spike Before the Diagnosis, the A1c can look normal while the first-hour glucose curve is already abnormal. In The Overflow Tank, BMI can look normal while fat is sitting in the liver, pancreas, muscle, and around the heart. In Miscalibrated Lipids, a standard lipid panel can look acceptable while ApoB, Lp(a), and ancestry-specific lipid architecture tell a different story.

The theme is the same:

The danger is assuming one test sees the whole phenotype.

The soft plaque problem

South Asians tend to develop cardiovascular disease earlier, often at lower BMI, often with diabetes or insulin resistance arriving sooner than expected, and often with risk calculators that underestimate what is happening inside the body.

The plaque phenotype may be different too.

A 2017 study in European Heart Journal – Cardiovascular Imaging compared 543 South Asian patients and 420 Caucasian patients with stable chest pain who underwent coronary CT angiography. South Asian patients were younger and had much higher diabetes prevalence. After adjustment, there was no major difference in total plaque burden or stenosis severity.

But the composition differed.

Among patients with atherosclerotic plaque, the proportion that was non-calcified was higher in South Asians: about 90.4% compared with 81.0% in Caucasian patients.

That is the key signal. Plaque type differed.

A newer 2025 matched CCTA study pushed this further. Researchers matched 165 South Asian patients to 165 non-Hispanic White patients by age, sex, BMI, and diabetes status. South Asians had higher total plaque percent atheroma volume, higher non-calcified plaque percent atheroma volume, and higher calcified plaque percent atheroma volume. The differences were especially visible in younger patients.

So when we say South Asians may carry more soft plaque, the claim is grounded in CCTA-based evidence.

This matters because CAC sees calcium, while CCTA can directly evaluate non-calcified plaque.

The Pakistani zero-CAC study that should make us pause

The cleanest patient-facing example comes from Pakistan.

In a 2024 Global Heart study, Shams and colleagues looked at 534 symptomatic patients of South Asian descent who had a zero Agatston calcium score and underwent CT coronary angiography.

These were symptomatic patients being evaluated for angina-like symptoms or angina equivalents.

The findings are exactly the kind of nuance we need:

• 28.8% had non-calcified, or soft, coronary plaque of some degree.

• 5.8% had obstructive coronary artery disease greater than 50%.

• Despite that, CAC=0 still carried a favorable long-term prognosis in the cohort.

That last point matters. The study made CAC=0 more precise:

CAC=0 can still be prognostically reassuring while failing to mean “no plaque.”

That is the conversation South Asian patients deserve: reassurance with precision.

It gave us more data for calibration rather than generalization.

If you are asymptomatic and low-risk, CAC=0 may be enough. If you are symptomatic, high-risk, or carrying a family history that makes everyone in the room uneasy, CAC=0 should leave room for another question.

What CCTA adds

Coronary CT angiography asks a different set of questions.

CAC asks whether plaque has calcified.

CCTA can show whether plaque is present at all. It can show where it sits, whether it narrows the artery, whether it is calcified or non-calcified, and whether there are features that make it more concerning.

That is why the 2021 AHA/ACC chest pain guideline gives CCTA a major role in evaluating intermediate-risk patients with acute or stable chest pain and no known coronary artery disease. In stable chest pain patients at intermediate-high risk, CCTA is effective for diagnosis, risk stratification, and guiding treatment decisions.

This is guideline-supported cardiology when the clinical context fits.

Where it becomes especially interesting for South Asians is the plaque-phenotype layer.

If our risk is partly driven by non-calcified plaque, the imaging test that sees non-calcified plaque may be the more biologically appropriate test in selected patients.

Again, selected patients.

A good pathway does not put every South Asian adult through a CCTA scanner. It reserves the stronger lens for the moments when CAC is being asked to answer a question it cannot fully answer.

Where Cleerly and HeartFlow fit

This is where the newer tools become important.

A standard CCTA read can tell us about stenosis, plaque, and anatomy. But AI-assisted quantitative plaque analysis can go further. Platforms like Heartflow and Cleerly-style AI-QCT can quantify total plaque volume, non-calcified plaque volume, low-attenuation plaque, and percent atheroma volume across the coronary tree.

That matters because plaque burden is not always obvious on a quick visual read, and because non-calcified plaque volume appears to carry prognostic information.

In the CERTAIN study, AI-QCT changed diagnosis or management in 57.1% of patients, reduced downstream testing by 37.1%, and increased preventive medication decisions compared with conventional CCTA interpretation alone.

In the CONFIRM2 registry, which followed 3,551 clinically indicated CCTA patients for a median of 4.27 years, non-calcified plaque volume and stenosis severity were independent predictors of major adverse cardiovascular events. Adding AI-QCT improved risk prediction beyond clinical models.

That is the Cleerly-type side of the story: what plaque is there, how much, and what kind?

HeartFlow and FFR-CT answer a different question: is a stenosis physiologically limiting blood flow? If CCTA shows an intermediate narrowing where anatomy alone does not settle the question, FFR-CT can help determine whether that lesion is causing vessel-specific ischemia and guide decision-making. NICE has also supported HeartFlow FFR-CT in selected patients with stable recent-onset chest pain who are already being offered CCTA.

Cardiac MRI belongs in this same functional lane. It is not the best tool for counting coronary plaque, but stress CMR is guideline-supported for diagnosing ischemia and estimating risk in intermediate-risk chest pain. It also adds tissue characterization: scar, myocarditis, cardiomyopathy, viability, and in some centers microvascular assessment.

These tools answer different questions. Cleerly-style AI-QCT is about plaque burden and plaque composition. HeartFlow-style FFR-CT and stress CMR are about physiology, ischemia, and downstream myocardial effects.

In the right patient, they are complementary. One tells us what is in the wall. The other tells us what it is doing to flow.

For South Asians, that combination may become particularly valuable because atherosclerosis can be earlier, quieter, and easier to miss before it becomes obviously obstructive.

How to Think About the Decision

Here is the practical way to think about it and organizing when CAC or CCTA might be an option.

Read more

This is the manifesto idea behind Zinda Futures: medicine is changing quickly, but every medical revolution carries the assumptions of the data used to build it.

The concept for the series came out of the recent South asian specific internal AI benchmark we ran at Zinda. We were testing whether AI systems could reason through South Asian clinical risk, and the result took me into a rabbit hole of understanding blind spots.

The benchmark made the core issue impossible to ignore: if the underlying clinical map is miscalibrated, every future tool built on top of it will inherit that miscalibration.

So great, once we have built out a refined understanding of the phenotype, how do we solve this?

The question is not only, “What is coming next?”

It should be temporal and intentional: will South Asian biology be included before the future becomes standard of care?

Why?

We are at a unique time in history where the the convergence of technologies in healthcare will accelerate in advances, shifting foundationally, which provides an opportunity to recalibrate South asian medicine from the ground up.

Most of us grew up with a version of medicine built around the disease label.

You have diabetes. You have high cholesterol. You have arthritis. You have coronary disease. You have fatty liver.

That label matters. But it is not the whole story.

The next era of medicine is moving one layer deeper. It will ask a different question:

Which pathway is failing in you?

For one person, the answer may be an inherited gene variant. For another, it may be an enzyme that does not work well, a transporter that changes how a medication is processed, an immune system that is overactivated, a mitochondrial system that cannot keep up, or a tissue environment that cannot repair properly.

That is where medicine is going. From diagnosis to mechanism.

The phrase I keep coming back to is mechanism-calibrated medicine. Or even calibrated medicine.

Not just personalized medicine in the loose consumer sense. Not just concierge care with more labs.

Mechanism-calibrated medicine means using population data, molecular biology, longitudinal measurement, imaging, genetics, bioenergetics, and clinical context together to ask: which mechanism matters in this person, and how should the interpretation be calibrated for the body in front of us?

And for South Asians, this matters more than most people realize.

The old model treated averages

Modern medicine has done incredible things. Vaccines, antibiotics, statins, insulin, surgery, transplantation. We should not pretend the old model failed.

But the old model was built around averages.

Average BMI cutoffs. Average diabetes thresholds. Average cardiovascular risk calculators. Average drug doses. Average trial participants.

The problem is that South Asians have often lived at the edge of those averages.

We develop insulin resistance at lower BMI. We can carry higher cardiometabolic risk even when standard numbers look “not that bad.” We have different patterns of visceral fat, muscle mass, lipid particles, inflammatory tone, and family-history risk. And in many clinical datasets, we are either underrepresented or grouped into broad categories that blur important differences between Indian, Pakistani, Bangladeshi, Sri Lankan, Nepali, and diaspora populations.

So when medicine becomes more personalized, the question is not just whether the technology gets better.

The question is whether the data finally includes us.

The new model treats pathways

Personalized medicine is not one thing. It is a stack of tools that all point in the same direction: treat the mechanism, not just the label.

That is why mechanism-calibrated medicine may be the better term for what is emerging.

It is not replacing population health. It is layering mechanisms on top of population data. The population tells us the prior probability: who tends to be missed, who develops disease earlier, which pathways cluster together, which thresholds need adjustment. The mechanism tells us what is actually happening inside the person: gene, enzyme, transporter, mitochondrial system, immune pathway, plaque phenotype, tissue niche, repair capacity.

Some examples are already here.

Clopidogrel, or Plavix, has to be activated by the CYP2C19 enzyme. If someone carries certain CYP2C19 loss-of-function variants, the medication may not work the way we expect. That is not a philosophical point. A 2023 JACC: Advances study in British Bangladeshi and Pakistani ancestry participants found a high burden of intermediate and poor CYP2C19 metabolizers, and poor metabolizers prescribed clopidogrel after myocardial infarction had a higher risk of recurrent MI.

Statins are another example. We often talk about statins as if dose is the only variable. But transporter genes such as SLCO1B1 and ABCG2 can affect exposure and tolerability for certain statins. That does not mean South Asians need less prevention. Usually the opposite is true: many need earlier and more aggressive ApoB lowering. But it does mean the path to staying on therapy may need to be smarter.

Lp(a) is another case we’ve spoken about at length on this publication.

Then there are enzyme-targeted treatments: enzyme replacement therapy, chaperone therapy, substrate reduction, mRNA therapies, and gene therapies for conditions where a specific enzyme pathway is broken. Fabry disease is a classic example, but the broader lesson is bigger than rare disease. The more precisely we understand a pathway, the more precisely we can intervene.

This is the pattern: genes, enzymes, receptors, transporters, immune signatures, RNA expression, and tissue repair pathways are becoming clinically actionable.

Some layers are already clinically actionable, like the ones above. Others, like mitochondrial and bioenergetic medicine, are emerging layers that may help explain why risk, repair, and treatment response differ across patients.

Regenerative medicine will be personal too

Regenerative medicine is exciting and something I believe will truly change human health trajectories. Yet, as a consequence of being nascent, it is often described too vaguely. People hear “stem cells” and imagine a universal repair fluid.

That is not how repair works.

A cell therapy or regenerative treatment enters a specific body. That body has a specific immune system, vascular supply, inflammatory state, mitochondrial capacity, glycation burden, microbiome, scar-forming tendency, and genetic background.

The tissue environment matters.

This is where genetics, metabolic, mitochondria and regenerative medicine start to merge. Induced pluripotent stem cells, or iPSCs, can be made from a patient’s own cells. In theory, those cells can be studied, corrected, engineered, differentiated into a target cell type, and potentially returned to the patient. This is still an evolving field, and safety matters: tumor risk, immune reaction, off-target editing, durability, and whether the new tissue behaves normally once it is back inside the body. All these must take into account the environment.

Regeneration will not just be “replace the tissue.” It will be: understand the patient’s repair biology, then decide what kind of repair is actually possible.

For South Asians, that question becomes even more important.

If a person has chronic insulin resistance, endothelial dysfunction, visceral adiposity, low-grade inflammation, vitamin D or B12 issues, mitochondrial stress, or accelerated vascular aging, those factors may affect how well tissue repair works.

Though I will premise this by saying we should be careful in specificying too much towards one phenotype as we do not yet have enough data, let alone South Asian-specific regenerative medicine data, to make sweeping claims.

But that is exactly the point.

If regenerative medicine is going to become personal, South Asian biology cannot be an afterthought.

The risk is a more sophisticated blind spot

There is a comforting story people tell about precision medicine: once medicine becomes genetic and molecular, it will automatically become fair.

I do not think that is necessarily the case.

We have seen this before. More technical medicine does not automatically become more equitable. Polygenic scores can fail when built on European-ancestry data. Pharmacogenomic algorithms can miss ancestry-specific variants. Pulse oximeters looked objective until they missed hypoxemia more often in darker skin. Kidney and lung equations used race as calibration and sometimes turned difference into delayed diagnosis. The lesson is simple: precision without representative data is not precision for everyone.

Even an AI model trained on incomplete clinical data can sound precise while being wrong for the person in front of it.

We have already seen the simpler version of this problem.

South Asian patients are told their BMI is fine when their visceral fat is not. Their LDL is “acceptable” when their ApoB or Lp(a) says otherwise. Their calcium score is reassuring when the question is soft plaque. Their glucose is normal until the post-meal spike appears. Their family history is noted but not modeled deeply enough.

The future can repeat the same mistake, just at a molecular level.

That is what we have to avoid.

What Zinda Futures is for

This is why Zinda Futures exists.

The AI benchmark was the trigger, but the lesson applies beyond AI.

Not to predict every new technology.

Not to chase hype.

But to ask a practical question before the future becomes standard of care:

What does South Asian health need to build now so the next generation of medicine is calibrated correctly from the beginning?

That means each future-facing question has to become a build agenda.

If medicine is moving toward pharmacogenomics, we need South Asian pharmacogenomic maps that clinicians can actually use.

If prevention is moving from LDL-C alone toward ApoB, Lp(a), and plaque imaging, we need South Asian-specific decision pathways for when the usual numbers are falsely reassuring.

If risk prediction is moving toward polygenic scores and AI models, we need to know which datasets trained them, which ancestry groups validated them, and where the scores fail.

If regenerative medicine is moving toward patient-specific repair biology, we need to study the South Asian tissue environment: insulin resistance, endothelial dysfunction, inflammation, mitochondrial stress, glycation burden, vitamin status, and the biology that determines whether repair works.

So Zinda Futures has a simple job:

• name where medicine is going

• identify the South Asian gap before it hardens into a standard

• explain the mechanism behind the gap

• turn the gap into something Zinda can measure, publish, build, or advocate for

How the Zinda framework compounds

The important thing is that Zinda Futures does not sit on top of vibes (sorry Andrej, I had to).

Read more

There is a sentence I have heard too many times in medicine:

“The numbers look fine.”

It is usually said with kindness. A clinician looks at the chart, sees a BMI in the low 20s, a fasting glucose that is only slightly high, an LDL that is not dramatic, a ten-year risk score that prints out …

Read more

It…

Read more

Read more

Read more

Feedback on topics always welcome!

Read more

Welcome back to Chai Shots, a series where we distill the latest science on South Asian health optimization.

In previous editions, we’ve been building a map of the South Asian metabolic phenotype. We’ve talked about the Fragile Engine (our smaller pancreatic beta-cell reserve), the Overflow Tank (our limited capacity to store fat safely), and the Adipone…

Read more

Read more

Read more

Goal remains the same…South Asian health optimization.

In Part I, we mapped the architecture of South Asian sarcopenia: the fetal programming that sacrifices muscle for fat, the dissociated mitochondria running hot, and the hormonal axes that compound it all. We established that muscle is not merely mechanical tissue but a metabolic organ.

The question now is what to do about it.

What follows is a comprehensive structure on how we can approach it.

The Intervention Architecture

If the South Asian body is built for conservation, storing fat, minimizing muscle, rationing energy, then the intervention strategy must be designed to override these defaults. It must be aggressive, multimodal, and sustained.

Consider how we approach dyslipidemia in South Asians which we touched a bit on Chai Shots earlier this month. We do not tell a patient with Lp(a) of 150 nmol/L to “just eat better.”

We recognize that the genetic architecture is set, that lifestyle alone, while necessary, cannot overcome the magnitude of the inherited risk.

We layer interventions: diet modification, exercise, and pharmacology working in concert. We monitor biomarkers. We titrate aggressively.

Sarcopenia in South Asians demands the same approach. The upregulated myostatin signaling suppressing hypertrophy, the developmental programming favoring fat over muscle, the mitochondria running hot but failing to clear lipids, these are not lifestyle problems with lifestyle solutions. They are genetic headwinds requiring metabolic intervention.

The current clinical approach, ”lift weights, eat protein,” is analogous to telling someone with familial hypercholesterolemia to reduce saturated fat. It is not wrong. It is woefully insufficient.

What follows is a comprehensive architecture built on three pillars:

• exercise programming (not generic, but South Asian-specific)

• peri-exercise optimization (what you do before and after matters as much as the training itself)

• pharmacological support (the interventions that address what lifestyle cannot).

This is not a menu of options. It is a layered framework that begins with non-negotiable foundations and builds upward toward targeted enhancement.

Foundation Layer: South Asian-Specific Exercise Programming

1. Volume and Frequency

Meta-analyses show resistance training at 70-85% of one-repetition maximum produces the greatest gains. However, for South Asians fighting myostatin-mediated suppression, the volume must exceed generic guidelines.

• Target: 3 - 4 sessions/week.

• The Reasoning: The signal must be louder to overcome the genetic noise. If you train infrequently, the conservation-programmed body simply reverts to baseline. You must present a constant demand that makes muscle maintenance a survival priority.

2. Compound Movements First

The largest glucose sinks in the body are the legs, glutes, and back. Squats, deadlifts, rows, and presses help build this as a core.

• The Reasoning: These muscles, when developed, create the metabolic runway we lack. Isolation exercises (like bicep curls) have their place, but they cannot substitute for the systemic anabolic signal that compound movements generate.

3. Time Under Tension & Eccentrics

Moving weight is not enough. The muscle must experience sustained mechanical tension to trigger the mTOR pathway.

• The Tactic: Controlled eccentrics (lowering phase), pauses at the bottom, and deliberate tempo.

• The Reasoning: Rushing through sets with momentum defeats the purpose. You must mechanically stress the tissue to signal to a conservation-programmed body: this expensive tissue is needed for survival.

4. Velocity Training (The Spark Plug)

Sarcopenia begins with motor neuron death, specifically the fast-twitch Type II neurons. If you never ask them to fire, they atrophy. Asians usually have significantly less Type II (fast-twitch) muscle fiber density compared to other populations..

• The Tactic: Move weights fast. The concentric phase (lifting) should be explosive. Consider “velocity days” with lighter loads (50-60% 1RM) moved with maximum speed.

• The Reasoning: Traditional “slow and controlled” lifting builds tension but misses the neural component. Explosive movement keeps the “spark plugs” (motor neurons) firing, preventing the neural disconnect that precedes muscle loss.

5. Capillary Density: The Last Mile

South Asian muscle often shows lower capillary density. You can flood the bloodstream with nutrients, but if the “last mile” infrastructure is sparse, the muscle starves.

• Intervention: Heat therapy (Sauna: 15-20 mins at 170-190°F, 3-4x weekly).

• The Reasoning: We have discussed the engine (mitochondria) and the fuel (lipid overflow), but not the road. Sauna use is one of the few interventions that stimulates angiogenesis (new blood vessel growth) independent of mechanical load, building the infrastructure for future training.

  • This reinforces the nitric oxide interventions discussed below, citrulline and beetroot are not merely for “the pump.” They are for vascular recruitment, opening collateral pathways to reach under-perfused tissue.

  • Additionally, heat therapy (sauna, 15-20 minutes at 170-190°F, 3-4 times weekly) is one of the few interventions that stimulates angiogenesis, the growth of new capillaries, independent of mechanical load. For the sarcopenic patient who cannot yet train intensely, sauna builds the infrastructure that will eventually support training.

6. Zone 2 for Metabolic Flexibility

By far my favorite…because it rewards patience and consistency. And I want to spend some time on the nuances here as Zone 2 spans a broad definition.

Resistance training alone does not clear the intramyocellular lipid trap. For that, you need aerobic work that preferentially burns fat.

Zone 2 cardio (heart rate 60-70% of maximum, conversational pace, 3-4 hours weekly) remains the foundation for this population. Lower baseline muscle mass and metabolic dysfunction make high-intensity work riskier, Zone 2 offers benefit without the same injury risk or stress load. More importantly, Zone 2 specifically trains fat oxidation pathways and is the only intensity that that upregulates MCT-1 transporters, the machinery that clears lactate. Since metabolically inflexible individuals often run with elevated baseline lactate, Zone 2 addresses this dysfunction directly in a way HIIT cannot. As capacity develops, higher-intensity work can be layered in.

The training stimulus requires nutritional support. Protein intake of 1.6-2.0 g/kg/day of ideal body weight, far higher than the 0.8 g/kg RDA designed for sedentary populations, is the threshold for supporting muscle protein synthesis in a body fighting innate headwinds geneticall. The lower recommendations (1.0-1.2 g/kg) are insufficient when starting from sarcopenic baseline.

Critical to understand: older or insulin-resistant muscle becomes “deaf” to protein signals. You need to exceed a threshold to trigger growth. Leucine, the amino acid that most potently activates the mTOR pathway governing muscle growth, must reach approximately 3.0g per meal to reliably stimulate muscle protein synthesis.

Here is the math based on traditional South Asian diets to provide sufficient protein for muscle building:

Food Source Amount Needed for 3g Leucine Whey protein isolate 1 scoop (25g) Chicken breast 120g (one palm-sized portion) Paneer 200g Greek yogurt 400g Dal (lentils) ~2.5 cups thick / ~9 bowls typical Rice + dal combined Still insufficient without fortification

The reality is that atypical bowl of dal contains roughly 0.5g of leucine which is not nearly enough. To hit the leucine threshold from dal alone, you would need to eat over 2.5 cups of thick lentils or nearly 9 bowls of typical homemade dal (almost a full cooker worth!!) in a single sitting. This is not a cultural judgment; it is a mathematical reality. For the vegetarian South Asians, supplementation is a necessity.

This is where cultural dietary patterns often collide with physiological need. Plant-based South Asian diets can provide adequate protein, but they require intention, combining rice with dal, emphasizing paneer and yogurt, adding whey or EAAs to meals. The casual vegetarianism that many families practice is not automatically sufficient for someone trying to build metabolic reserve.

Peri-Exercise Optimization: What Happens Before and After

Read more

There is a scene that repeats itself in clinics across the world, in SF, NYC, in Leicester, in Dubai, in Bangalore, so often that it has become almost invisible. A South Asian man, let us call him Raj, sits on an examination table. The paper beneath him crinkles. The fluorescent light hums. The doctor, who is usually kind, usually competent, usually pre…

Read more

Read more

If you’ve been following lipid management guidelines, you’ve probably heard about PCSK9 inhibitors, the “miracle drugs” that cut LDL-C by 50-60%. Repatha, Praluent, and the newer siRNA agent inclisiran have transformed cardiovascular prevention.

So here’s the uncomfortable question to ask: What if this entire therapeutic strategy was optimized for the wr…

Read more

I…

Read more

Read more

On the Substack, we’ve spent the last few weeks building out a map of the South Asian metabolic phenotype with beta cells, ghrelin, mitochondrial efficiency, genotypes.

Today we add another layer: adiponectin.

You’ll need some chai for this as this will be the most in-depth piece we have done so far.

Read more