Home · Services · Sankofa Cohort
SANKOFA

Cohort

Medicine tested on the people it will treat.

A laboratory scientist loading blood samples into an analyser

Why it matters

1.1%of the world’s clinical trials were hosted in Africa in 2023, home to nearly one person in five

Medicines approved on results from a narrow population reach 1.5 billion people whose responses were barely measured. Efavirenz and warfarin show the cost: standard dosing guides can give people of African descent the wrong dose.

What Sankofa Cohort does

Sankofa Cohort brings African evidence to global medicine. Pharmaceutical companies, contract research organisations (CROs), research institutions and hospitals build custom cohorts from longitudinal African records, test their hypotheses against them and create synthetic data to design and power clinical trials. CROs use the same cohorts to test and validate trial designs, endpoints and models. Regulators, including the US FDA, expect trial populations to reflect the people a medicine will treat, and Sankofa supplies African evidence at a depth few can match. Sankofa’s models can also train on site through federated learning: only the results, the model weights, are shared between institutions, and the data remains local and sovereign. In a study across 20 hospitals, models trained this way performed 38% better at hospitals they had never seen than models trained at a single site.

SANKOFACohortHospitalData stays localUniversityData stays localPharmaceutical companyData stays localResearch instituteData stays localModel trains at each institutionOnly the weights are sharedShared weights, stronger models; the data remains local and sovereignModel trains at each institutionOnly the weights are sharedHospitalData stays localUniversityData stays localPharmaceutical co.Data stays localResearch instituteData stays localSANKOFACohortShared weights, stronger models;the data remains local and sovereign

Sources: WHO clinical-trial data for 2023, via the World Economic Forum, July 2025; Dayan and others, Nature Medicine, 2021, federated learning across 20 institutions.

In detail

What it covers

For pharmaceutical companies, CROs, research institutions and hospitals.

  • Custom cohortsGroups of people defined by condition, treatment, risk and response
  • Validation cohortsReady-made and custom cohorts for CROs to test and validate trial designs, endpoints and models
  • Synthetic trial dataRepresentative synthetic populations to design and power clinical trials
  • Trial feasibilityWhere to find the right people for a clinical trial, and how many
  • Real-world evidenceHow medicines and treatments perform in everyday care
  • Safety surveillanceEarly signals of side effects and resistance
  • Health economicsThe cost of care and the value of new treatments
  • Federated learningModels trained at each institution, with only the weights shared
An analyst reviewing outcome charts at her desk in the evening

For global medicine

African evidence for every stage of a medicine’s life.

Pharmaceutical companies

From the first hypothesis to trial design, recruitment and safety monitoring after approval, evidence of how a medicine performs across the widest range of people on earth.

Contract research organisations (CROs)

Diverse, well-characterised cohorts from longitudinal African records to test and validate trial designs, endpoints and models, with feasibility and recruitment planning for the trials CROs run.

Work with us

Let’s talk.

Whether you build AI models, develop new medicines, run health services, fund health programmes or invest, we’d like to hear from you.

General enquiries: hello@sankofa.health
Media: press@sankofa.health
LinkedIn: Sankofa Health ↗

Sankofa Health (Pty) Ltd
Suite 102, 25 Richefond Circle
Ridgeside Office Estate, Umhlanga Rocks 4319
South Africa

Sankofa Health (Pty) Ltd uses the details you send only to answer your enquiry and keep in touch about it. We do not add you to any mailing list. You may ask us at any time to correct or delete your details by writing to privacy@sankofa.health.