Harnessing the Power of Global Health Studies for Sickle Cell Disease: Validation of a Rapid, Open-Source, Paper-Based Screening Assay in a Cohort of 1103 Tanzanian Children

Introduction: Global health research is hampered by the lack of inexpensive and reliable assays to annotate clinical study cohorts for geographically localized endemic genetic disorders. The most prevalent modifier in Africa is sickle cell disease (SCD, HbSS). In SCD, sickling of erythrocytes by polymerization of hemoglobin S (HbS) causes vaso-occlusion and hemolysis, leading to significant morbidity/mortality and constituting a major healthcare and economic burden. The inhomogeneity bias introduced into the clinical data of patient cohorts by the changes associated to undiagnosed SCD is a significant challenge. Thus, our aims were to: a) optimize a paper-based screening test for SCD for its application in suboptimal and resource-limited conditions, b) demonstrate the feasibility of retrospective SCD annotation of a large cohort of children enrolled in a global health research study, and c) extract novel evidence on potential differences in clinical presentation between febrile children with and without SCD enrolled in this cohort.

Materials & Methods: First, we adapted and simplified a paper-based SCD assay (Piety et al. 2017) for suboptimal sample conditions (frozen EDTA-whole blood) and limited-resource settings. Five microlitres of thawed EDTA-whole blood was solubilized (high phosphate buffer with saponin/sodium hydrosulfite), and blotted on standard blotting paper. Dots formed by capillarity, and their aspect depended on HbS%. A high-resolution scanner and two smartphone brands were compared for image capture. An algorithm for visual calling of the presence an HbS allele was developed, and the freeware ImageJ determined peripheral/central pixel intensities. An HbS dosimetric curve was determined after identification of optimal parameter correlations by scatter plot matrix. Second, blood samples from 1103 children aged 9 to 35 months enrolled in a randomized open prospective non-inferiority pediatric cohort study on algorithm-based management of febrile illnesses in Tanzania (ePOCT Trial, 2014-6, NCT02225769, Swiss/Tanzanian IRB approval, Keitel et al.2017) were assayed with the optimized paper test. In parallel, confirmation by gold-standard alkaline Hb electrophoresis (Helena Biosciences, UK) was performed. Third, clinical data from the newly SCD-annotated ePOCT database was analyzed (STATA software suite) to identify clinical features associated with febrile illnesses in the HbAA/AS/SS groups.

Results: All EDTA-blood samples frozen onsite in Tanzania (shipped and then stored for >1 year) could be assayed. The test was rapid (5'), inexpensive (0.05 USD/test) and accessible (widely available reagents, no requirement for costly equipment, open-source freeware). Determination of HbS status (homozygous HbSS/AA or heterozygous HbSA) was effective both visually and by informatic means. Smartphone cameras yielded identical results (R²=0.96 & 0.97) when compared to a high-definition scanner. Local prevalence was at 1.6% and 13.6% for HbSS and HbAS respectively, in line with previously published data. By visual sorting of blood dots, the assay was 100% sensitive and 75.8% specific for the determination of HbSS-SCD and 97.6% sensitive and 87.5% specific for the identification of an HbS allele (in an hetero- or homozygous state). After analysis of the annotated ePOCT multiparametric clinical dataset, we observed that cough as the main complaint was more prevalent in children with HbSS (OR 3.1, 95% CI: 1.1-8.3). HbSS children were more severely malnourished (MUAC/age average z-scores -0.7 vs 0.0, p=0.01) and slightly more anemic (Hb 9.1 vs 9.8 g/dL, p=0.06) when compared to HbAA counterparts. Interestingly, children with HbAS had a better nutrition status than those with HbAA (MUAC/age average z-score 0.14, p<0.01). Also, education of mothers of HbSS children was significantly lower (primary education OR 0.3 (95% CI: 0.1-0.8).

Conclusions: We demonstrated that the retrospective annotation of clinical cohorts for SCD with an optimized paper-based assay is feasible, straightforward and inexpensive. Our approach has the potential to eliminate the interpretation bias associated with SCD, and thus facilitate the downstream analysis of valuable datasets in other studies. As a proof of principle, we examined a properly annotated dataset highlighting novel features of febrile illness in children with SCD and HbAS carriers in Africa.