Improving diagnosis of non-malarial fevers in Senegal: Borrelia and the contribution of tick-borne bacteria

Abstract

The worldwide decline in malaria incidence is revealing the extensive burden of non-malarial febrile illness (NMFI), which remains poorly understood and difficult to diagnose. To characterize NMFI in Senegal, we collected venous blood and clinical metadata from febrile patients and healthy controls in a low malaria burden area. Using 16S and unbiased sequencing, we detected viral, bacterial, or eukaryotic pathogens in 29% of NMFI cases. Bacteria were the most common, with relapsing fever Borrelia and spotted fever Rickettsia found in 15% and 3.7% of cases, respectively. Four viral pathogens were found in a total of 7 febrile cases (3.5%). Sequencing also detected undiagnosed Plasmodium, including one putative P. ovale infection. We developed a logistic regression model to distinguish Borrelia from NMFIs with similar presentation based on symptoms and vital signs. These results highlight the challenge and importance of improved diagnostics, especially for Borrelia, to support diagnosis and surveillance.

Figure 1:

Overview of burden of malarial and NMFI In Thiès, Senegal and NMFI pathogens identified In 2019 a. Patients suspected of malaria across 28 health facilities in the Thiès region (data provided by Programme National de lutte contre le Paludisme); gray shading indicates periods of sample collection for this study, b. Types of pathogens detected in febrile patients in dry and rainy seasons in 2019. Plasmodium detected by RDT or RNA-mNGS, bacterial infections detected by 16S amplicon sequencing, and viral infections detected by RNA-mNGS. c. Viral infections detected by RNA-mNGS d. Bacterial infections detected by 16S sequencing and e. Co-infections detected in febrile patients from 2019.

Figure 2:

Characterization of qPCR-confirmed Borrelia and RDT-confirmed Plasmodium across both years a. Borrelia, Plasmodium, and non-Borrelia NMFI (Other febrile) incidence across dry and rainy seasons in 2018 and 2019. b. Age distribution of Borrelia, Plasmodium, and other febrile patients, c. Maximum likelihood phylogenetic tree (midpoint rooted) of IGS sequences from this study (red) in the context of available reference sequences for B. crocidurae (purple), B. duttonii (cyan), and B. recurrentis (green). For B. crocidurae isolates, location of collection indicated in label and host species indicated with node shape as available.

Figure 3:

Clinical characteristics across both years of Borrelia (n = 39), Plasmodium (n = 75), and other febrile (n = 412) patients Including a. symptoms reported by patients, b. exposures reported by patients, and c. vital signs. For 2019 only, d. blood cell counts In Borrelia (n = 17), Plasmodium (n = 41), other febrile (n = 146), and healthy (n = 104) e. Serological profile of a subset of patients with Borrelia (n = 9), Plasmodium (n = 11), viral Infection (n = 10) compared to healthy patients (n = 10); cytoklnes/chemoklnes with at least one significant difference between groups shown [see Supplementary File 1 for results for all cytomkines/chemokines tested]. *Mann-Whitney-Wilcoxon test two-sided, p-value annotation legend: ns: p <= 1.00e+00, *: 1.00e-02 < p <= 5.00e-02, **: 1.00e-03 < p <= 1.00e-02, ***: 1.00e-04 < p <= 1.00e-03, ****: p <= 1.00e-04

Figure 4:

Results of best performing weighted logisistic regression model to distinguish Borrelia infection from other NMFI using clinical data, including demographics, symptoms, exposures, and vital signs (trained on 2018–2019 data, tested on the same data with bootstrapping and cross-validation, n = 451). a. Model performance metrics and b. odds ratios for all features retained In the final model after feature selection.