Study area and population
This cross-sectional study, nested within a cohort study, was conducted in May–June 2022 and included 990 participants aged 18 to 88 years living in 21 villages near Sibiti, Republic of Congo. Participants had been previously screened for Loa microfilaremia [20]. For the current study, individuals with blood microfilariae detected in 2021 were matched for sex and age (± 5 years) with two amicrofilaremic individuals from the same village. Detailed description of the study population has been published previously [8, 21].
The sample size for the underlying cohort study (n = 990) was calculated to provide 80% statistical power to detect an increased risk of malaria (1.4-fold) and pneumonia (2.4-fold) among microfilaremic compared to amicrofilaremic individuals, based on the hypothesis that loiasis may impair splenic function and thereby increase susceptibility to infections. However, due to the lack of prior data, the sample size was not calculated to evaluate differences in renal outcomes.
Participation in the study required informed consent from all individuals. The research was approved by the Ethics Committee of the Congolese Foundation for Medical Research (036/CIE/FCRM/2022) and by the National administrative authorities of Congo (376/MSP/CAB/UCPP-21).
Sociodemographic and exposure covariables
Data on age, sex, anthropometric measurements (weight, height, and body mass index—BMI), and tobacco use were collected from each participant. Blood pressure was measured supine after 10 min rest. High blood pressure (HBP) was classified into three stages (see Table 1) and mean arterial pressure (MAP) was defined as (frac{{{text{Systolic BP + 2 }} times {text{Diastolic BP}}}}{{3}}).
Laboratory procedures for parasitic infections
Fifty µl of capillary blood were collected by finger-prick between 10 a.m. and 4 p.m. (to take into account the day and night fluctuation of Loa MFD) to prepare thick blood smears (TBS) which were then Giemsa-stained, and examined at 100 × magnification by experienced technicians to count Loa and Mansonella perstans mf. Each TBS was read twice; the arithmetic mean was recorded. Prior exposure to Onchocerca volvulus was assessed using the antibody detecting Onchocerciasis rapid test targeting Ov16 antigen (Drugs & Diagnostics for Tropical Diseases, San Diego, USA). Two skin snips were taken from seropositive individuals using a 2 mm Holth punch, incubated in saline at room temperature for 24 h and examined microscopically to count the emerged mf. O. volvulus MFD was calculated as the arithmetic mean of the two counts, expressed per snip.
Schistosoma haematobium infection was investigated in participants with haematuria detected by urine dipstick. Positive cases underwent urine filtration, Lugol staining, and microscopic examination for eggs. Soil-transmitted helminths (STH) were identified via microscopic examination of morning stool samples, transported within 6 h and processed immediately or after overnight storage at 6 °C. Stool smears were prepared using the Kato-Katz method and examined at 40 × magnification. For asymptomatic Plasmodium infection, thin blood films from venous blood (heparinized tube) were stained with RAL 555 (RAL Diagnostics, Martillac, France) and examined microscopically. An indirect ELISA was used to quantify the Immunoglobin G (IgG) antibodies levels against P. falciparum antigens [22].
Biological examinations
Creatinine levels were measured for each patient in whole blood with a point-of-care device (iSTAT-1; Abbott Point of Care, Princeton, NJ, USA). For logistic reasons, glycated haemoglobin (Hb1Ac) and fasting serum total cholesterol level, triglycerides, HDL, and LDL were only measured in a random subset of patients, using a point-of-care device (Afinion 2, Abbott Rapid Diagnostics, Bièvres, France). Eosinophilia and lymphocyte counts were conducted using the HemoCue WBC DIFF system (HemoCue AB, Ängelholm, Sweden).
Arterial stiffness assessment
Arterial stiffness was assessed through finger-toe pulse wave velocity (PWV) using the pOpmètre device (Axelife SAS, Paris, France), which records pulse waves via two infrared photodiode sensors at the finger and toe. This method is a validated, non-invasive alternative to carotid-femoral PWV, which requires skilled staff and femoral artery access [23].
Outcome: chronic kidney disease
Ultrasound examination
The US examination of the kidneys was performed by experienced physicians (VD, LR) using a CX-50 device (Philips Medical Systems, Suresnes, France). Measurements included height, width, length. A second reading of the digital images (DICOM files) was performed to confirm and characterize kidney parenchymal or structure abnormalities (pyelocaliceal dilatation, cystic, or other description).
Urine collection and dipstick analysis
Each participant provided a morning midstream urine sample, following instructions from a trained nurse. Urinalysis was performed immediately using dipsticks (Reactif 10SL Urinalysis Strips, 41101-M, Nal von Minden, Moers, Germany) to detect proteinuria and haematuria. A single physician interpreted the results and did a second confirmatory dipstick if the initial result was positive. No discrepancies were observed. Haematuria was assessed as negative, traces, + , + + , or + + + . Positive cases (≥ 1 +) underwent urine filtration and microscopic examination for S. haematobium eggs. Proteinuria was classified as negative, traces (< 0.3 g/l), level–1 (0.3–1 g/l), level–2 (1–3 g/l), or level–3 (> 3 g/l). Albumin-to-creatinine ratio (ACR, mg/mmol) was measured for proteinuria ≥ level–1 using the Afinion 2 device.
Estimated glomerular filtration rate (eGFR) and chronic kidney disease (CKD) classification
Given limited data on creatinine distribution and CKD in our population, we applied two formulas. First, the 2009 CKD-EPI equation without ethnicity factor, as current recommendation [24, 25]. Second, the EKFC equation, a newer formula more robust across all ages [26, 27]. Only one published Q-value (median normal creatininemia) exists for Central Africa [28], based on 494 urban individuals [25]. Therefore, we estimated our own Q-values by calculating median creatinine levels in individuals without HBP or renal abnormalities (proteinuria or haematuria). CKD-EPI and EKFC eGFRs were expressed in ml/min per 1.73 m2, and classified per KDIGO (Kidney Disease: Improving Global Outcomes Guidelines) categories [29]. Renal abnormalities (RAb) were defined as ACR > 3 mg/mmol or ≥ 1 + haematuria, excluding ultrasound-confirmed causes (9 cases of pyelocaliceal dilatation suggestive of lithiasis were excluded, as the associated haematuria was most likely of urological rather than nephrological origin and thus not indicative of intrinsic renal disease).
Statistical analysis
Continuous variables with normal distribution were presented as mean ± standard deviation (SD); non-normal variables were reported as median (interquartile range). Explanatory variables included: age (continuous), sex (male/female), tobacco use (yes/no), MAP (continuous), PWV (≥ 12 vs. < 12 m/sec), lymphopenia (< 1200/µl), hypereosinophilia (> 1500/µl), Ascaris lumbricoides and/or Trichuris trichiura (present/absent), Loa MFD (0, 1–7999, 8000–19 999, ≥ 20 000 mf/ml), anti-Plasmodium falciparum IgG (quartiles), and Plasmodium smears (negative/positive). BMI was not included since anthropometric parameters are integrated into eGFR estimations. Lipid profiles and diabetes were not included due to substantial missing data (see Results). The PWV cutoff (12 m/sec) reflects its established association with organ damage [30]. HIV status was unknown, but lymphopenia < 1200/µl was used as a proxy.
In our study, 25% weighed < 50 kg (10% < 45 kg, range: 30–120 kg), 25% had a height < 157 cm (10% < 152 cm, range: 141–190 cm), and 25% had a BMI < 18.9 kg/m2 (10% < 17.8, range: 13.2–47.7). Given this heterogeneity, we chose to apply Dubois body surface area (BSA) correction to both CKD-EPI and EKFC equations to better assess associations. eGFR was calculated with CKD-EPI and then de-indexed using individual BSA via the Dubois formula: BSA = 0.007184 × Weight(kg)0.425 × Height(cm)0.725, resulting in CKD-Dubois and EKFC-Dubois values (ml/min).
First, a saturated linear regression model on eGFR was performed. Because KDIGO-CKD staging could not be used directly in the regression analysis—due to the necessity of adjusting for BSA and the small number of participants in advanced CKD stages, which limited statistical power—we defined three alternative severity thresholds based on the distribution of eGFR values in our study population to investigate a potential gradient effect of CKD severity. These were the median (Cutoff-1), the 10th percentile (Cutoff-2) – corresponding to the global prevalence of CKD –, and the 2nd percentile (Cutoff-3) –reflecting the prevalence of advanced CKD. Thresholds were calculated separately for each equation. Given that the presence of RAb is an essential component of CKD staging, it was incorporated in the categorization and thus the categories were: “ ≥ Cutoff-1 & no RAb”, “ ≥ Cutoff-1 & RAb”, “Cutoff-1 – Cutoff-2 & no RAb”, “Cutoff-1 – Cutoff-2 & RAb”, “Cutoff-2 – Cutoff-3 & no RAb”, “Cutoff-2 – Cutoff-3 & RAb”, “ < Cutoff-3 & no RAb”, “ < Cutoff-3 & RAb”.
Due to the ordered nature of outcome variables, ordinal logistic models were initially considered. However, they were rejected after a Brant test indicated a proportional odds violation. Instead, multinomial logistic regression was used. A manual stepwise backward selection (P < 0.100) was applied to the saturated model via the likelihood ratio test. To prevent convergence failure, Loa MFD was re-categorized into three groups (0, 1–19 999, ≥ 20 000 mf/ml). Finally, potential interactions between Loa MFD, age, sex, and eosinophilia were assessed using the likelihood ratio test. After finalizing the model, we extracted risk probabilities. We then estimated the population attributable fraction (PAF) from a logistic model including the final significant variables identified (age, sex, and Loa microfilaremia as a binary variable) using the KDIGO CKD classification with the EKFC formula.
All statistical analyses were performed using Stata 18 (StataCorps LP, College Station, Texas, USA).