The Evidence Base
4,321 sources. One picture of South Asian health.
Standard medicine reads South Asian patients against reference ranges built for other populations. The result is risk that hides in plain sight: diabetes at normal BMI, heart attacks at forty, kidneys declining while the labs read fine. This is the evidence that makes it visible, assembled from the primary literature and growing daily.
Plus 1,325 case reports under review. The picture sharpens every day.
01The framework, live
Two layers, honestly graded. Layer 1 mechanisms are universal biology engaged at South Asian-relevant thresholds. What makes the framework South Asian is Layer 2, the integration of genetics, diet, culture, and diaspora context. Each concept shows its current study count and evidence maturity.
02We count the evidence against us
Every study is graded by what it does to the framework: supports it, qualifies it, reframes it, or contradicts it. A concept with nothing but support is treated as under-examined, not proven. Honest medicine counts its counterevidence.
View as table
| Concept | support | neutral | qualify | reframe | contradict | Ungraded |
|---|---|---|---|---|---|---|
| The Fragile Engine | 117 | 57 | 6 | 23 | 2 | 69 |
| Burn and Crash | 97 | 55 | 5 | 8 | 5 | 70 |
| The Overflow Tank | 126 | 27 | 5 | 11 | 2 | 49 |
| The Undersized Filter | 42 | 29 | 0 | 2 | 1 | 41 |
| Nutrient Substrate | 36 | 21 | 2 | 5 | 2 | 8 |
| The Signal Fire | 27 | 13 | 1 | 2 | 1 | 12 |
| The Thrifty Engine | 28 | 6 | 1 | 4 | 0 | 9 |
03Where the evidence sits in a life
The framework models four stages of phenotype evolution. Standard care sees nothing until Stage 3. The evidence concentrates exactly where patients look normal.
04A young field, moving fast
Studies by publication year. Most of what medicine knows about South Asian metabolic health was published in the last three years; the field is accelerating, and the historical record is being reviewed systematically against the complete published literature.
05The benchmark: what generic AI misses
Sixty South Asian clinical vignettes, six scored dimensions, run against frontier models and CDS products. The same model, given the Zinda framework as context, jumps double digits. The knowledge is the bottleneck, not the model.
Internal benchmark, temperature 0, blind prompts that never mention South Asia. Judge calibration and inter-rater reliability are pre-publication work in progress; treat scores as directional. Test cases are never published, so the benchmark stays uncontaminated.
| Platform | Cases | SA-CDS Score |
|---|---|---|
| Claude Opus 4.6+Zinda | 60 | 91.4 |
| Gemini 3.1 Pro+Zinda | 58 | 85 |
| OpenEvidence | 31 | 80.4 |
| GPT-5.4+Zinda | 60 | 72.2 |
| Gemini 3.1 Pro | 60 | 72 |
| Kimi K2.5 | 60 | 69.4 |
| GLM 5.1 | 60 | 68 |
| DeepSeek V3.2 | 60 | 67.9 |
| GPT-5.4 | 60 | 65.3 |
| Claude Opus 4.6 | 60 | 65.1 |
06The N=1 repository
3,016 published case reports of South Asian patients, aggregated from literature that large cohorts ignore, filed across 17 clinical domains. Individually anecdotes; together, the pattern library.
07Recently added, and what it means
The newest landmark findings in the evidence base, each one a reason the framework exists.
- Predicting complications for diabetes in South Asians: Beyond convention.Conventional complication-prediction tools for type 2 diabetes are derived from European-ancestry cohorts and traditional clinical variables, perform suboptimally in South Asians, and leave substantial residual risk unexplained. The review surveys non-conventional predictors — li PMID 42161723
- Prevalence of undiagnosed type 2 diabetes in South Asia: A systematic review and meta-analysis.Pooled prevalence of UNDIAGNOSED type 2 diabetes in the South Asian general adult population was 6.67% (95% CI 4.99-8.58%), with extreme heterogeneity (I2=98.5%) and a very wide prediction interval (0.33-19.75%). Country estimates: Sri Lanka 30.54% (17.90-44.82), Nepal 7.02% (4.2 PMID 42541604
- Differential Pathophysiological Drivers of Susceptibility to Type 2 Diabetes and Metabolic Dysfunction-Associated Steatotic Liver Disease: Ethnic Differences in Insulin Dynamics, Whole-Body Fat Metabolism, and Organ-Specific Lipid Deposition.A three-way ancestry comparison (South Asian vs African Caribbean vs White European) that lands on the 'palette' model of type 2 diabetes: elevated risk in both minority groups, but via *different* phenotypes. South Asians: greater subcutaneous and liver fat, more severe insulin PMID 41698854
- Multi-ancestry polygenic risk scores for the prediction of type 2 diabetes and complications in diverse ancestries.Single-ancestry PRSs performed best in European (incremental AUC 0.07-0.14) and East Asian (0.02-0.16) ancestries and POORLY in South Asian (0.02-0.04), African/African American (0.02-0.03) and Admixed American (0.02-0.04) ancestries, tracking GWAS sample size. Multi-ancestry PRS PMID 42061389
- Linkage disequilibrium and allelic heterogeneity explain variation in coronary artery disease risk at 9p21 across populations and reduced effect in Africans.The first and most replicated genome-wide-significant coronary artery disease locus, 9p21.3, was robust in European, East Asian, Middle Eastern and Admixed American ancestry groups, only NOMINAL in South Asians, and absent in Africans. The African non-replication is not due to ab PMID 42385718
08Access
The observatory is public. The engine's depth is tiered: registration is free and instant, collaborator access is granted personally. Long-form writing lives on the Zinda Substack, one list, no spam.
- This observatory, updated daily
- The framework, honestly graded
- Selected case highlights
- Chai Shots and Field Notes
- Browse all 3,016 case summaries
- Per-concept evidence lists with sources
- Benchmark methodology in full
- Coverage ledger detail
- Clinical parameter registry with provenance
- Compiled lens protocols
- Gene Atlas entries
- Full study dossiers