Large-scale modelling of β-thalassaemia carrier prevalence: geostatistical predictions and population estimates from a global evidence base

Background

Evidence on β-thalassaemia carrier prevalence is fragmented across heterogeneous surveys and national summaries, leaving within-country variation and geographic concentration of carriers poorly resolved. This study characterises β-thalassaemia carrier prevalence and carrier populations at high spatial resolution and identifies where additional data would most improve geographic coverage.

Methods

Using carrier-prevalence records curated through ITHANET, a global haemoglobinopathy resource, Bayesian geostatistical modelling generated high-resolution predictions across 25 countries. Predictions were integrated with population data to estimate national and subnational carrier numbers. Prediction uncertainty and gaps in existing survey coverage informed prioritisation of healthcare sites for additional surveys.

Findings

Predicted prevalence varied markedly within countries, and highest-prevalence areas did not coincide with the largest carrier populations. National mean prevalence was highest in Cyprus (12·99%), whereas India, China, and Pakistan had the largest carrier populations, together accounting for 81·21% of estimated carriers. Random five-fold cross-validation favoured a model capturing broad geographic patterns and location-specific variation. Predictive performance deteriorated when geographically clustered records were withheld together, indicating reduced reliability across larger gaps in geographic sampling and the need for additional data from poorly characterised areas. Survey prioritisation identified 183 healthcare sites across 17 countries for future carrier-prevalence surveys.

Interpretation

By distinguishing areas of high carrier prevalence, large carrier populations, and evidence gaps, this updatable framework provides a subnational evidence base to inform β-thalassaemia prevention planning and guide future data collection, screening, and genetic counselling.

Funding

COST Action HELIOS, Horizon Europe HemaFAIR, Horizon 2020 GenoMed4ALL, and the Cyprus Institute of Neurology & Genetics.

Research in context

Evidence before this study

Carrier identification is central to β-thalassaemia prevention because the geographic distribution of carriers is a fundamental determinant of where at-risk reproductive pairings can occur. Population surveys and screening programmes have established the broad distribution of β-thalassaemia and informed prevention strategies in several high-prevalence settings, but the underlying evidence varies substantially in sampling frame, diagnostic method, period of observation, and geographic coverage. Carrier prevalence reflects the distribution of inherited HBB variants across populations and is shaped by population history and structure, including endogamy, admixture, and migration; substantial variation can therefore occur within national boundaries. National estimates are useful for broad epidemiological assessment but cannot resolve such local heterogeneity, whereas individual surveys provide direct evidence only for the populations and locations sampled. Geostatistical modelling can integrate geographically dispersed observations to estimate carrier prevalence in unsampled intervening locations while quantifying uncertainty. This approach has enabled high-resolution characterisation of the spatial distribution of other haemoglobinopathies, including global mapping of sickle haemoglobin allele frequency with estimates of neonates with sickle cell anaemia or sickle cell trait, mapping of haemoglobin C allele frequency across Africa, and estimation of α-thalassaemia variant frequencies in Thailand. Comparable large-scale spatial modelling of β-thalassaemia remains limited and geographically restricted. Consequently, the fine-scale distribution of β-thalassaemia carrier prevalence within countries, and the extent to which these patterns translate into the geographic distribution of carrier populations, remain insufficiently characterised.

Added value of this study

To our knowledge, this study provides the first high-resolution characterisation of β-thalassaemia carrier prevalence across 25 countries with sufficient curated carrier-prevalence evidence for spatial modelling. Quality-appraised carrier-prevalence records were integrated within a Bayesian geostatistical framework to generate predictions at approximately 5 km × 5 km resolution, with uncertainty quantified across the modelled domain. Alternative model formulations were compared, and ecological, demographic, and healthcare service availability covariates were evaluated for their contribution to predictive performance. Carrier-prevalence predictions were combined with count-preserving population surfaces to estimate the number and spatial distribution of carriers. Prediction uncertainty and existing sampling coverage were then used to identify healthcare sites where additional epidemiological surveys could most improve the geographic coverage of the empirical evidence. This framework therefore distinguishes areas where β-thalassaemia carrier prevalence is highest, where carrier populations are largest, and where additional data are most needed to improve spatial inference, while allowing estimates and survey priorities to be updated as new carrier prevalence data become available.

Implications of all the available evidence

High-resolution β-thalassaemia carrier-prevalence adds a missing epidemiological layer to prevention planning by showing where screening is likely to have the highest yield, while carrier numbers indicate where downstream requirements for confirmatory diagnosis and genetic counselling could be greatest. Used alongside existing programme data, these measures can help distinguish settings in which established prevention strategies remain appropriately configured from those in which their geographic reach, intensity, or capacity may warrant reassessment, without implying that new programmes are necessarily required. The priority site analysis extends this framework by identifying healthcare sites where additional surveys could strengthen geographic sampling coverage the most. The resulting maps and data have been made publicly available through the ITHANET portal for interactive exploration and download, broadening access to fine-scale evidence to support policy and programme decisions.