Non-Communicable Diseases
- Home
- Disciplines
- Non-Communicable Diseases
Analysis support for the cardiovascular, metabolic, cancer and mental health research that Africa's epidemiological transition demands.
Non-communicable diseases account for a rapidly rising share of mortality across the continent, and the research base needed to respond to them is thin. Most risk prediction models, polygenic scores and treatment guidelines in use were developed in populations that look nothing like the ones African clinicians serve.
That gap is a research opportunity as much as a problem, and it is analysis-intensive work: large cohorts, long follow-up, complex confounding, and models that must be validated locally before they can be trusted.
Research Areas We Support
Analysis capability applied across the NCD landscape.
Cardiovascular Disease
Risk factor analysis in cohort and cross-sectional studies, hypertension and stroke epidemiology, validation of risk scores such as Framingham and SCORE2 in African populations, and blood pressure trajectory modelling.
Diabetes and Metabolic Disease
Glycaemic outcome analysis, metabolic syndrome prevalence and clustering, gestational diabetes studies, and the atypical diabetes phenotypes reported across sub-Saharan Africa that fit poorly into existing classifications.
Cancer
Tumour genomic and transcriptomic analysis, survival and competing risks modelling, cancer registry analysis, and molecular subtyping of cancers with distinctive African epidemiology — cervical, oesophageal, prostate, breast and Kaposi sarcoma.
Chronic Respiratory Disease
COPD and asthma cohort analysis, spirometry data handling and the reference equation problem, and analysis of air pollution and biomass fuel exposure as risk factors.
Mental Health
Analysis of validated screening instrument data, psychometric assessment of instruments translated and adapted for local populations, and longitudinal modelling of outcomes.
Sickle Cell and Haemoglobinopathies
Genotype-phenotype association analysis, survival and crisis frequency modelling, newborn screening programme evaluation, and modifier gene studies.
Analysis Challenges We Handle
Risk models that do not transfer
A cardiovascular risk score calibrated on a European cohort will systematically miscalibrate on an African population, in ways that vary by age and sex. We assess calibration explicitly, recalibrate where the data allows, and report where a model simply should not be used.
The same applies to polygenic risk scores, which lose most of their predictive power when transferred across ancestry groups. We are direct about that rather than reporting a score as though it were portable.
Long follow-up and messy data
NCD cohorts run for years and the data reflects it — changing instruments, staff turnover, loss to follow-up, competing mortality. We handle this explicitly:
- Multiple imputation and sensitivity analysis for missing data, with assumptions stated
- Competing risks models where death from other causes is common
- Mixed-effects and GEE models for repeated measurements within individuals
- Time-varying exposures and covariates
- Joint models for longitudinal biomarkers and time-to-event outcomes
Multi-omics integration
Where studies combine genomics, transcriptomics, metabolomics and clinical data we integrate them rather than analysing each in isolation — and we are honest about the sample sizes required before integration adds anything beyond noise.
