Chai Shots #010: South Asians Are Barely Visible in the Trials That Shape Heart Care
A review of major heart trials found that South Asians were barely counted and often lost in the reporting.
Welcome back to Chai Shots, our series reviewing research that should be changing South Asian health.
With any dataset, as a physician, I want to know two things: where are we starting, and where are we trying to go?
That starting point matters in South Asian health. Someone living in the West may have a very different experience from someone living on the subcontinent. Even within the diaspora, a family in London may not have the same experience as a family in Bay Area.
Diet, environment, access to care, migration history, income, and stress can all differ. So can the way a health system classifies someone.
Research involving the South Asian diaspora remains sparse. And recruitment is only one part of the problem. A trial may enroll South Asians, then publish the results under a category too broad to tell us anything useful about them.
A recent paper measured how often this happened in major cardiovascular trials.
Let’s get into it.
What the research found
A systematic review published in JACC: Asia examined 310 randomized cardiovascular trials published between 2018 and 2022 in The New England Journal of Medicine, JAMA, and The Lancet. Those trials enrolled 1,036,737 people.
South Asians made up 3.2% of the participants. Nearly two-thirds of them came from one quality-improvement trial conducted in Kerala. Remove that trial and representation falls to 1.1%.
The diaspora figure was worse. Across 150 trials with 386,887 participants in North America, Europe, and Oceania, the researchers could identify only two South Asian participants.
Only two South Asians were identifiable in regions where millions of us live and where cardiovascular disease is common in our families.
The reporting problem continued after recruitment. Thirty-five trials clearly enrolled South Asians, 19 reported how many, and six reported their post-randomization outcomes. Those outcome data are what allow us to see what happened after treatment began.
Other studies placed South Asians inside the broad category “Asian.” That may simplify a table, but it cannot tell us whether a result held for an Indian, Pakistani, Bangladeshi, Sri Lankan, Chinese, Korean, or Japanese patient. This is something that is changing and improving, but older studies have nearly no granularity here.
The authors found no identifiable South Asian participants in the reviewed trials of heart failure, cardiac arrest, or valvular heart disease. So they cant tell us at all, how South Asians responded in those trials.
The limits of standard practice
Standard practice inherits the limits of its data. When a population is barely included, or included and then hidden during reporting, the direct evidence for that population remains weak.
Many cardiovascular mechanisms are shared across human beings. ApoB-containing particles enter the artery wall. High blood pressure damages vessels. The clinical setting can still differ. South Asians may develop diabetes earlier, carry more visceral fat at a lower BMI, have less lean mass, or carry an elevated Lp(a). Family history, diet, migration, and access to care add more context.
A recommendation may still be correct, but we should be honest about how directly it was tested on the population we apply said recommendation to.
We have discussed parts of this problem in One in Four, Miscalibrated Lipids, and The Calcium Score Can Be Zero. The Plaque Risk May Not Be.. A useful measure can still miss something important about the person sitting in front of you.
The same gap reaches, and magnifies, with medical AI. In What Happens When Medical AI Sees South Asian Patients?, the systems we tested could repeat that South Asians carry higher cardiometabolic risk. They struggled to let that fact change the decision when a generic guideline said the patient looked normal.
No model can recover a South Asian subgroup result that researchers never published.
What 1.1% does not tell us
The 1.1% figure belongs to this cardiovascular review. It covered three journals and five publication years. Pregnant and pediatric populations were excluded, and inconsistent reporting may have hidden participants the reviewers could not identify.
This paper did not measure representation in kidney disease, cancer, fatty liver disease, neurology, autoimmune disease, mental health, women’s health, pediatrics, or pharmacogenomics. We cannot carry 1.1% into those fields and present it as fact. and yes, it is possible that representation may be higher in other areas where South Asians were studied more deeply. A good example could be treatments that are more aligned with Ayurvedic or Eastern medicine, which sometimes have more South Asian data available.
Each field needs its own audit. We still do not know how large the research gap is across the rest of medicine.
The review also cannot prove that a standard therapy works differently in South Asians. Underrepresentation leaves us with less certainty about treatment effects; whether the effects differ remains unanswered.
The Zinda Imperative
When we started Zinda, it was a response to how little health guidance reflected the realities of South Asians, especially those living in the West. We were working inside a medical system largely calibrated on other populations.
That led us to build the N=1 Case Repository, which now contains roughly 3,300 case reports and case series. We are also building a South Asian Gene Atlas, research papers, an Evidence Library, and a clinical framework focused on the places where generic thresholds can create false reassurance.
For us, an AI-first approach means finding and organizing evidence that already exists but remains scattered or overlooked. We want to use that work to identify better questions for future studies and trials. We cannot afford to ignore existing signals while another one or two generations wait for the ideal evidence base.
what that then allows us to do is model out disease using tools such as digital twins and simulators and eventually in silico modeling for South Asians. That accelerates clinical application both for current treatments, but more importantly for emerging treatments.
We will publish and test our findings. The model and research infrastructure remain proprietary.
Randomized trials remain necessary. Our work can help show where those trials are overdue and what they should ask. This is part of the work behind Zinda Futures.
Physicians and patients can also act on the uncertainty that exists now. When a medication or recommendation is offered, patients should feel comfortable asking:
Were people like me included in the evidence behind this recommendation? If they were, what did the data show for them?
Whatever the answer, the uncertainty should be visible.
— Omar
Zinda Health is a reader-supported publication. To receive new posts and support this work, consider becoming a free or paid subscriber.