Non-linear effects of parasite density on symptom burden and hematologic indices in pediatric Plasmodium falciparum malaria

1 Introduction

Malaria remains a leading cause of morbidity and mortality globally, with Plasmodium falciparum responsible for the most severe and life-threatening forms of the disease. In 2022, an estimated 249 million malaria cases and 608,000 deaths were reported worldwide, with over 95% occurring in sub-Saharan Africa, disproportionately affecting children under five years of age (Schofield and Grau, 2005; Marsh and Kinyanjui, 2006; Doolan et al., 2009). The clinical spectrum of malaria ranges from asymptomatic parasitemia to severe disease characterized by multi-organ dysfunction, reflecting a complex interplay between host immunity, parasite virulence, transmission intensity, and environmental factors (Clark et al., 1987; Kwiatkowski, 1990; Luxemburger et al., 1998).

Parasite density is widely regarded as a key determinant of malaria severity. High parasite loads are associated with amplified systemic inflammation, endothelial activation, microvascular sequestration, and impaired tissue perfusion, contributing to clinical manifestations such as high fever, prostration, anemia, metabolic acidosis, and cerebral malaria (Koram et al., 2003; Ndungu et al., 2005; O’Meara et al., 2008; Mackinnon and Marsh, 2010). However, the relationship between parasite density and clinical presentation is not strictly linear. In high-transmission settings, repeated exposure leads to partial immunity, allowing children to tolerate moderate or even high parasite densities with minimal symptoms (D’Alessandro et al., 1995; Price et al., 2001; Reyburn et al., 2004a). Conversely, in non-immune or semi-immune individuals, relatively low parasite densities may provoke acute febrile illness and systemic inflammatory responses.

Several studies from holoendemic African regions have attempted to define parasite density thresholds associated with fever or clinical malaria, though reported thresholds vary widely by geography, season, age, and immune status (Menendez et al., 2000; Ekvall, 2003; Williams, 2006). For example, Ndungu and colleagues demonstrated that parasite densities above approximately 1,000 parasites/µL were strongly predictive of febrile episodes among Kenyan children, while lower-density infections still contributed to mild disease and sustained transmission (Ndungu et al., 2005). Such heterogeneity highlights the limitations of assuming a uniform or linear relationship between parasite density and clinical manifestations, and underscores the need for analytical approaches capable of capturing threshold effects and non-proportional changes across the parasite density spectrum.

Hematologic abnormalities are hallmark features of P. falciparum malaria and reflect both parasite-driven and host-mediated mechanisms. Anemia is among the most common complications, particularly in children, arising from destruction of infected and uninfected erythrocytes, bone marrow suppression, dyserythropoiesis, and inflammatory cytokine-mediated inhibition of erythropoiesis (Newton et al., 2000; Verhoef et al., 2002; Crawley, 2004; Wassmer et al., 2004; Okebe et al., 2014). Clinically, malaria-associated anemia contributes to fatigue, poor feeding, delayed neurocognitive development, and increased mortality when hemoglobin levels fall below critical thresholds (Rodríguez-Morales et al., 2005; McMorran et al., 2009; Lacerda et al., 2011). Nevertheless, the relationship between parasite density and anemia severity is complex and often weak in cross-sectional analyses, as anemia in endemic settings reflects cumulative parasite exposure, nutritional deficiencies, hemoglobinopathies, and co-existing infections such as helminthiasis or HIV (Facer, 1994; Kho et al., 2018).

Thrombocytopenia is another frequent hematologic abnormality in malaria and often develops early in the course of infection. Platelet counts below 150 × 10⁹/L have been reported in up to 70–80% of patients with acute malaria (Ogutu et al., 2003; Erhart et al., 2004; Keating et al., 2010). Proposed mechanisms include splenic sequestration, immune-mediated platelet destruction, direct parasite–platelet interactions, and consumption driven by endothelial activation and microvascular inflammation (Sowunmi et al., 1997; Armah et al., 2005; Taylor et al., 2008; Fendel et al., 2010). Importantly, platelet counts typically recover rapidly following parasite clearance, making thrombocytopenia a sensitive and dynamic marker of acute parasitemia (Srichaikul et al., 1988; Kotepui et al., 2020; Kotepui et al., 2022). Several studies have therefore proposed platelet count as a useful adjunct for differentiating malaria from other febrile illnesses in endemic settings (Kwiatkowski et al., 1990; Schofield et al., 1993; Smith et al., 1994; Snow et al., 1996).

In contrast, white blood cell (WBC) counts in uncomplicated malaria are often normal or mildly reduced, with inconsistent associations with parasite density (Felger et al., 1994; Casals-Pascual et al., 2006; Prentice et al., 2017). Observed alterations likely reflect immune redistribution phenomena, including lymphopenia, monocytosis, or neutropenia in more severe disease (Bejon et al., 2007; Roca-Feltrer et al., 2008a; Rogerson and Carter, 2008). Among routinely available hematologic parameters, platelet count therefore appears more responsive to fluctuations in parasite burden than hemoglobin or WBC.

In many malaria-endemic, resource-limited settings, rapid and reliable quantification of parasite density is not routinely available at the point of care. Although microscopy remains the diagnostic gold standard, its effectiveness is frequently constrained by limited laboratory capacity, variable quality assurance, and delayed turnaround times (Moody, 2002; Reyburn et al., 2004b; Bell et al., 2006). Rapid diagnostic tests (RDTs) facilitate parasite detection but do not provide quantitative estimates. Consequently, clinicians often rely on clinical symptoms and basic laboratory indices—such as hemoglobin, platelet count, and WBC—for triage and risk stratification, particularly in primary and secondary healthcare facilities.

Earlier studies examining relationships between parasite density and clinical or hematologic outcomes have largely employed linear regression or categorical parasite density groupings (Greenwood et al., 1991; Beare et al., 2006; Hanson et al., 2012). While informative, such approaches may obscure biologically meaningful thresholds and underestimate non-linear or saturable responses. Many biological processes underlying malaria pathogenesis—including cytokine release, fever induction, platelet activation, and endothelial dysfunction—exhibit non-linear dynamics rather than simple dose–response relationships (Ashley et al., 2018; World Health Organization, 2023). Conventional linear models may therefore fail to capture critical inflection points across the parasite density continuum (Hopkins et al., 2020).

Flexible non-linear modeling approaches, such as natural cubic splines, provide a data-driven framework for characterizing complex exposure–response relationships without imposing linear assumptions (Chotivanich et al., 2000; Buffet et al., 2011). These methods allow identification of density ranges where symptom burden or hematologic abnormalities change sharply, offering clinically actionable insights. However, to date, relatively few studies in sub-Saharan Africa have comprehensively applied spline-based or other non-linear modeling techniques to evaluate how parasite density relates simultaneously to symptom burden and hematologic profiles in pediatric malaria, despite the region bearing the highest global disease burden (Clark et al., 2006).

We therefore hypothesized that (1) parasite density exhibits non-linear associations with clinical symptom burden; (2) thrombocytopenia correlates more strongly with parasite density than hemoglobin or WBC; and (3) these relationships reveal biologically meaningful thresholds relevant for clinical risk stratification. Accordingly, this study applied natural cubic spline modeling within generalized regression frameworks to evaluate associations between parasite density, symptom burden, and routine hematologic indices among Nigerian children presenting with acute P. falciparum malaria. Findings from this work have implications for early triage, clinical decision support, and malaria surveillance in resource-limited, high-transmission settings.

2 Methodology

2.1 Study design and setting

This investigation was a cross-sectional analytical study conducted at Ajikobi Cottage Hospital, Ilorin, Kwara State, Nigeria. The facility serves as a first-level health center for surrounding urban and peri-urban communities and receives a high volume of pediatric malaria cases, particularly at first clinical presentation. As such, the setting was well suited for capturing routine, real-world patterns of symptomatology and hematologic abnormalities among children presenting with acute malaria, rather than community-level prevalence or longitudinal disease trajectories.

The study adhered to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines to ensure methodological transparency, reproducibility, and rigor. All assessments were carried out at baseline (Day 0, Hour 0), prior to initiation of anti-malarial therapy, to minimize treatment-related confounding and to reflect clinical decision-making at initial presentation.

2.2 Study population and recruitment

Participants were recruited consecutively from the pediatric outpatient clinic of the hospital during routine clinic hours. Children presenting with symptoms suggestive of malaria were screened for eligibility, and 357 participants with microscopy-confirmed Plasmodium falciparum infection were enrolled. Consecutive recruitment was employed to reduce selection bias and to reflect typical presentation patterns in a high-transmission, resource-limited clinical setting.

Although the study was conducted at a single health facility, the sample size is comparable to several hospital-based pediatric malaria studies in endemic settings and was sufficient to support flexible non-linear modeling of parasite density–outcome relationships (Folarin et al., 2021). Nonetheless, the study population primarily represents symptomatic children presenting for care and may not fully capture asymptomatic or community-level malaria burden.

2.3 Inclusion criteria

Participants were included if they were within the pediatric age range as defined by the hospital’s service, presented with clinical features consistent with malaria, had confirmed P. falciparum parasitemia on thick or thin smear microscopy, and had complete baseline clinical and laboratory data, including hemoglobin, platelet count, and white blood cell count.

2.4 Exclusion criteria

Participants were excluded if they had received anti-malarial treatment for the index illness prior to presentation, had known chronic hematological disorders or comorbidities that could distort laboratory results (e.g., aplastic anemia, leukemia, or other hematologic malignancies), required urgent resuscitation that precluded study procedures, or if consent/assent could not be obtained.

Recruitment was performed during routine clinic hours by trained study staff to reflect typical presentation patterns. To minimize selection bias, no prescreening beyond initial clinical triage was performed.

2.5 Ethical considerations and informed consent

The study protocol was reviewed and approved by the University of Ilorin Teaching Hospital Ethics and Research Committee (UITH ERC) and the Kwara State Ministry of Health Ethics Committee. The study complied fully with the Declaration of Helsinki and national research ethics regulations.

All Informed Consent Forms (ICFs) and Participant Information Sheets (PIS) were administered in English and local languages, depending on participant preference. The consent process included clear explanations of the study objectives, procedures, potential risks, expected benefits, and confidentiality safeguards. Parents or legal guardians provided written informed consent, and assent was obtained from children who were of appropriate age and capacity. Signed consent forms were stored in secured, access-controlled cabinets, and all data were de-identified prior to analysis.

2.6 Clinical assessment

Upon enrollment, trained clinicians conducted a standardized evaluation using a structured case report form. Data collected included demographic characteristics, vital signs (notably axillary temperature), and a comprehensive 21-item symptom inventory covering common malaria-associated clinical manifestations, including headache, vomiting, abdominal pain, loss of appetite, cough, fatigue, diarrhea, and other systemic or localized symptoms.

Each symptom was graded for intensity on a 0–3 severity scale using a standardized clinical grading guide, and symptom duration was recorded in days. These data were used to derive two summary measures of clinical presentation at enrollment:

• Symptom count – the total number of distinct symptoms reported by each participant, capturing the breadth of clinical involvement.
• Symptom severity index – the sum of individual symptom grades, representing overall symptomatic burden at presentation.

Symptoms were weighted equally by design. This approach was chosen to provide a transparent and reproducible summary of clinical burden without imposing subjective assumptions regarding the relative importance of individual symptoms. In the absence of a validated pediatric malaria symptom weighting system applicable across endemic settings, equal weighting allows the severity index to function as a descriptive composite measure of contemporaneous symptom burden, rather than as a diagnostic or prognostic severity score.

All clinical assessments were completed prior to the availability of laboratory results to minimize observer bias and to ensure that symptom grading reflected clinical presentation independent of parasitological or hematologic information.

2.7 Laboratory procedures

2.7.1 Malaria diagnosis and parasite density determination

Venous or capillary blood was collected at presentation for preparation of thick and thin blood films, which were stained and examined according to standard operating procedures aligned with WHO guidelines. Parasite density was quantified using one of two site-standardized approaches: counting against 200 red blood cells (RBC) or 500 white blood cells (WBC). The unit used for each patient was documented and included as a covariate in multivariable analyses to account for potential measurement differences. Parasite counts were transformed using the natural logarithm (log_density) to reduce skewness and stabilize variance for statistical modeling.

2.7.2 Hematology

Hemoglobin concentration, platelet count, and white blood cell count were measured at baseline using the hospital’s routine hematology analyzer. The laboratory adhered to internal quality control protocols, and data were extracted electronically to avoid transcription errors. No post-hoc recalibration was applied.

2.7.3 Data management and quality assurance

All clinical and laboratory data were entered into a secure REDCap/Excel database and double-entered to detect entry errors. Rigorous range and logic checks were applied to identify implausible values, including biologically impossible or extreme laboratory measurements. Any discrepancies were verified against source documents. Missing data were minimal and largely limited to item-level symptom nonresponse. A complete-case approach was used for primary analyses, and sensitivity checks confirmed that exclusion of missing observations did not materially affect effect estimates or statistical inference. All personal identifiers were removed prior to analysis, and only de-identified datasets were used in the statistical workflow.

2.7.4 Exposure, outcomes, and covariates

The primary exposure of interest was log-transformed parasite density (log_density). Clinical outcomes included symptom count, symptom severity index, and the occurrence of any symptom with a severity grade of ≥3. Hematologic outcomes included anemia severity (modeled as an ordinal variable according to hemoglobin thresholds), any hematologic grade ≥3 event, and thrombocytopenia (any) based on site reference standards. Continuous hemoglobin, platelet, and white blood cell counts were modeled to characterize linear and non-linear trends across parasite densities. Axillary temperature and smear density unit were prespecified as adjustment covariates because of their clinical and measurement relevance. This transformation was applied to reduce left-skewness and to improve numerical stability in regression models. Axillary temperature and smear density unit (RBC- vs WBC-based counting) were prespecified adjustment covariates because of their biological relevance to symptom expression and their potential influence on parasite density estimation.

2.8 Statistical analysis

Associations between parasite density and clinical or hematologic outcomes were examined using generalized regression models incorporating natural cubic splines to allow for flexible, non-linear exposure–response relationships without imposing linearity assumptions.

For each outcome, parasite density entered the model as a spline function of log-transformed density Equation 1:

(1) ɡ(E[Outcomei])=β0+f(log(Yi))+XiTβ

where:

• Outcome_i denotes the outcome of interest for participant i (e.g., symptom count, symptom severity index, severe symptom status, anemia severity, thrombocytopenia, or continuous hematologic indices),
• E[Outcome_i​] = the expected (mean) value of the outcome for participant i, given parasite density and covariates.
• g(.) is the appropriate link function (log for count models, logit for binary/ordinal models, identity for linear models),
• Y_i​ denotes the parasite density for participant i (parasites/µL), and log(Yi) is the log-transformed parasite density used in modeling,
• f(.) is a natural cubic spline function (3 degrees of freedom) capturing the non-linear effect of log parasite density,
• X_i is the vector of adjustment covariates (axillary temperature and smear density unit), and
• β is the vector of regression coefficients for covariates.
• Symptom count was modeled using quasi-Poisson regression to account for overdispersion, with results expressed as incidence rate ratios (IRRs).
• Symptom severity index was analyzed using negative binomial regression to accommodate overdispersion in summed severity scores.
• Binary outcomes—including presence of any symptom with severity grade ≥3, thrombocytopenia (any), and any hematologic grade ≥3 event—were modeled using logistic regression, with results expressed as odds ratios (ORs).
• Anemia severity, categorized according to hemoglobin thresholds, was modeled using ordinal logistic regression under the proportional odds assumption, which was assessed using score tests.
• Continuous hematologic indices (hemoglobin, platelet count, and white blood cell count) were analyzed using linear spline regression with three segments, allowing slopes to vary across parasite density ranges.

All models were adjusted a priori for axillary temperature and smear density unit. Robust (heteroskedasticity-consistent) standard errors were used to improve inference reliability. Statistical significance was assessed using two-sided p-values <0.05. Model discrimination for thrombocytopenia was evaluated using the area under the receiver operating characteristic curve (AUC) with 95% confidence intervals.

All statistical analyses were performed in R (version 4.1).

2.9 Ethical approval and data protection

The study received full ethical approval from the University of Ilorin Teaching Hospital Ethics and Research Committee and the Kwara State Ministry of Health Ethics Committee (UITH/HREC/CAT/189^B/342). Written informed consent was obtained from parents or guardians, with child assent when appropriate. Confidentiality was protected through secure storage of consent documents, restricted access to identifiable information, and de-identification of all analytic datasets. Electronic data were password-protected, and paper records were stored in locked cabinets accessible only to authorized personnel.

No protocol deviations occurred, and all analyses adhered to the pre-specified plan. The study design, rigorous consent procedures, robust laboratory and statistical methods, and strong ethical oversight together ensure the integrity, reproducibility, and validity of the findings.

3 Results

3.1 Participant characteristics and symptom profile

A total of 357 children aged 6 months to 12 years with microscopy-confirmed Plasmodium falciparum malaria were included in the analysis. The median age was 6 years (IQR: 4–8 years). Most participants presented with mild disease manifestations at first clinical contact (Day 0, Hour 0). The most common symptoms were headache (22.4%) and vomiting (22.1%), followed by abdominal pain (5.3%) and loss of appetite (4.5%) (Table 1). Gastrointestinal and respiratory symptoms were less frequent, and systemic symptoms such as fatigue and myalgia occurred in ≤1% of participants. Neurologic, cutaneous, psychiatric, and urinary symptoms were rare or absent at presentation. The median severity grade for most symptoms was 1 (mild), with fatigue and myalgia persisting slightly longer (median duration 3.5 and 4 days, respectively) (Table 1).

Table 1. Baseline prevalence, severity, and duration of malaria-associated symptoms at presentation among enrolled children (n = 357).

Values represent symptom prevalence at enrollment (Day 0, Hour 0) prior to initiation of anti-malarial therapy. Symptom severity was graded using a standardized 4-point clinical scale (1, mild; 2, moderate; 3, severe; 4, life-threatening). Median severity grade is reported for participants reporting each symptom. Symptom duration (days) was obtained by caregiver or participant report. Symptoms with zero observed prevalence are listed collectively; severity grade and duration are not applicable (—). Overall, the majority of reported symptoms were mild in intensity and of short duration at presentation.

3.2 Parasite density and symptom burden

The association between parasite density and symptom burden at presentation was examined using non-linear spline models (Table 2). These analyses were designed to characterize cross-sectional relationships observed at first clinical contact, rather than temporal changes or prognostic thresholds. Figure 1 illustrates the modeled probability of presenting with at least one severe symptom (grade ≥ 3) across the distribution of parasite density at enrollment. The probability of severe symptoms was low at lower parasite densities and increased at higher densities, with a steeper rise observed beyond approximately 4,000 parasites/μL. This pattern indicates a concentration of severe clinical phenotypes among children presenting with higher parasite burdens, rather than a transition over time from mild to severe disease.

Table 2. Logistic regression analysis of the association between parasite density and the presence of any severe symptom (grade ≥ 3) at presentation.

Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using logistic regression with parasite density modeled as a natural cubic spline of log-transformed density (three degrees of freedom). Spline terms represent different regions of the parasite density distribution (lower, intermediate, and upper density ranges) and should be interpreted jointly rather than individually. Models were adjusted for axillary temperature and smear density unit. The presence of any symptom with severity grade ≥3 served as the binary outcome. The upper parasite density range demonstrated a strong but imprecise association with severe symptom presentation, reflecting limited events at extreme density levels.

Overall, children presenting with higher parasite densities had greater odds of severe symptoms at the time of assessment, although confidence intervals widened at extreme densities, reflecting limited numbers in these ranges.

Children with higher parasite densities showed markedly greater odds of presenting with severe symptoms (grade ≥ 3), though confidence intervals were wide at very high densities.

3.3 Parasite density and symptom count

Total symptom count at presentation was also evaluated in relation to parasite density using a quasi-Poisson regression model with natural cubic splines (Table 3). This analysis assessed whether children with higher parasite burdens tended to report a broader range of symptoms at the same time point, without implying symptom accumulation over time. A non-linear association was observed, with minimal variation in symptom count across low and intermediate parasite densities, and an upward shift in the upper density range. The third spline term showed borderline statistical significance (IRR = 2.75; 95% CI: 1.02–7.87; p = 0.052), indicating that children presenting with higher parasite densities tended to report a greater number of concurrent symptoms.

Table 3. Quasi-Poisson regression analysis of the association between parasite density and total symptom count at presentation.

Incidence rate ratios (IRRs) and 95% confidence intervals (CIs) were estimated using quasi-Poisson regression to account for overdispersion in symptom counts. Parasite density was modeled as a natural cubic spline of log-transformed density with three degrees of freedom. Spline terms correspond to different regions of the parasite density distribution and should be interpreted collectively. Models were adjusted for axillary temperature and smear density unit. Results indicate an upward trend in symptom count at higher parasite densities, with the strongest effect observed in the upper density range.

These findings suggest that higher parasite burdens are cross-sectionally associated with broader symptomatic expression, rather than reflecting progressive symptom development.

Figure 2. Spline plot of symptom count vs. parasite density, showing non-linear increase in burden at high densities.

3.4 Parasite density and symptom severity index

The symptom severity index, representing the summed severity of reported symptoms at presentation, was analyzed using negative binomial regression with spline terms (Table 4). This outcome captures overall symptomatic burden at a single time point, rather than disease progression. A strong association was observed in the upper parasite density range, with the third spline term showing a statistically significant increase in severity index (IRR = 5.08; 95% CI: 1.77–15.8; p = 0.003). In contrast, lower and intermediate density ranges showed no clear association with severity index.

Table 4. Negative binomial regression analysis of the association between parasite density and symptom severity index at presentation.

Incidence rate ratios (IRRs) and 95% confidence intervals (CIs) were estimated using negative binomial regression to accommodate overdispersion in the summed symptom severity index. Parasite density was modeled as a natural cubic spline of log-transformed density with three degrees of freedom. Spline terms represent different regions of the parasite density distribution and should be interpreted jointly. Models were adjusted for axillary temperature and smear density unit. The upper parasite density range was associated with a statistically significant increase in symptom severity index, indicating greater symptomatic burden at higher densities.

These results indicate that children presenting with higher parasite densities tend to exhibit disproportionately greater symptom severity at presentation, consistent with intensified host inflammatory responses at higher parasitemia. The observed spline pattern reflects cross-sectional heterogeneity in clinical presentation and should not be interpreted as a severity threshold or predictive marker of subsequent clinical deterioration.

3.5 Visualization of vital sign trends and symptom clustering

Figures 3a–e show smooth spline relationships between parasite density and vital signs (temperature, heart rate, respiratory rate, systolic and diastolic blood pressure). Although most vital parameters showed weak linear trends, heart rate and temperature demonstrated slight upward trajectories with increasing parasite density. Supplementary Figure 1 shows a symptom correlation heatmap. The co-occurrence of headache, vomiting, and abdominal pain was common, forming the dominant cluster of presenting symptoms. Neuropsychiatric and urinary symptoms were rare and largely uncorrelated with other symptom domains.

3.6 Parasite density and anemia severity

Ordinal logistic regression with natural cubic splines was used to evaluate the association between parasite density and anemia severity categories (none, mild, moderate, severe). As shown in Table 5, no statistically significant dose–response relationship was observed across parasite density levels at the 5% significance threshold. The odds ratios for spline terms 1–3 were 0.74 (95% CI: 0.14–3.82; p = 0.73), 0.89 (95% CI: 0.00–199; p = 0.91), and 0.60 (95% CI: 0.12–3.09; p = 0.54), respectively. All confidence intervals included the null value of 1.0, and corresponding p-values exceeded 0.05, indicating no statistically significant association between parasite density and anemia severity at presentation. Similarly, axillary temperature (OR 0.86; 95% CI 0.68–1.10; p = 0.23) and smear density unit per 500 WBC (OR 1.40; 95% CI 0.70–2.82; p = 0.34) were not significantly associated with anemia severity.

Table 5. Ordinal logistic regression analysis of the association between parasite density and anemia severity at presentation.

Odds ratios (ORs), 95% confidence intervals (CIs), and p-values were estimated using proportional odds (ordinal logistic) regression. Parasite density was modeled as a natural cubic spline of log-transformed density with three degrees of freedom. Models were adjusted for axillary temperature and smear density unit. No statistically significant associations were observed at the 5% level.

The absence of statistical significance likely reflects several factors. First, anemia in endemic settings is multifactorial and may not correspond directly to parasite density measured at a single time point. Second, the wide confidence intervals—particularly for the intermediate spline segment—suggest limited precision in certain density strata, potentially due to sparse observations in extreme categories. Third, hemoglobin levels may reflect cumulative or prior parasitemia rather than contemporaneous parasite load.

Figure 4 illustrates the distribution of hemoglobin concentrations at presentation, which appears approximately symmetric with mild dispersion across anemia categories. Figure 5 demonstrates a shallow downward spline gradient across increasing parasite densities; however, substantial overlap of confidence intervals reinforces the absence of a statistically significant density–hemoglobin relationship.

Collectively, these findings indicate that anemia severity at presentation is not independently predicted by parasite density within this cohort.

Figure 4 illustrates the distribution of hemoglobin concentrations at presentation, which is approximately symmetric with a slight leftward tail, with most values clustering between 7 and 11 g/dL corresponding predominantly to mild-to-moderate anemia categories within the cohort. Severe anemia was relatively uncommon, and there was no evidence of a distinct bimodal or threshold pattern across hemoglobin strata. The distribution suggests substantial variability in hemoglobin levels that is not tightly aligned with a single dominant density range.

Figure 5 illustrates the spline-modeled relationship between log-transformed parasite density and hemoglobin concentration. Although the fitted spline curve shows a modest downward trajectory across increasing parasite densities, the slope is shallow and the 95% confidence bands overlap substantially across the density spectrum. There is no clear inflection point or density threshold at which hemoglobin declines sharply. The absence of a pronounced curvature or steep gradient reinforces the ordinal regression findings, indicating that contemporaneous parasite density does not independently predict anemia severity at presentation within this cohort. Together, these graphical findings support the interpretation that hemoglobin levels reflect heterogeneous and potentially cumulative processes rather than a direct, dose-dependent response to parasite density measured at a single clinical encounter.

3.7 Parasite density and hematologic grade ≥ 3 events

Logistic regression analysis assessing the odds of any hematologic grade ≥ 3 event (e.g., severe anemia, thrombocytopenia, leukopenia) revealed no statistically significant associations with parasite density (Table 6).

Table 6. Logistic regression of any hematologic grade ≥ 3 vs. parasite density.

Logistic regression results for severe hematologic outcomes vs. parasite density. No significant associations were observed. CI, confidence interval; OR, odds ratio.

Spline term ORs were 0.70 (95% CI: 0.07–14.2; p = 0.80), 417 (95% CI: 0.02–2.88 × 10⁸; p = 0.30), and 1.71 (95% CI: 0.04–54.6; p = 0.80) for terms 1–3, respectively. The wide confidence intervals at higher densities indicate instability of estimates in that range, with no clear evidence of a consistent trend.

3.8 Continuous hematologic indices across parasite density

Linear spline models were fitted to further explore continuous relationships between log-transformed parasite density and hemoglobin, platelet count, and white blood cell (WBC) count.

• Hemoglobin (Hb): Non-significant downward trends were observed across spline segments (Table 7), consistent with the expected impact of parasite burden on red cell destruction.
• Platelet counts: There was a significant positive shift in platelet counts across the first two spline segments, with flattening at higher densities.
• WBC: No clear relationship was found between parasite density and total white blood cell count.

Table 7. Linear spline models for continuous hematologic indices vs. parasite density.

Linear spline model coefficients for hemoglobin, platelet count, and WBC vs. log-transformed parasite density. Trends were strongest for platelet counts, especially in the lower-to-mid density ranges.

3.9 Parasite density and thrombocytopenia

A strong and statistically significant association was observed between parasite density and risk of thrombocytopenia (platelet count <150 × 10⁹/L). As shown in Table 8, the odds ratios were 9.47 (95% CI: 1.51–67.0; p = 0.018) for spline 1 and 8.99 (95% CI: 1.54–56.7; p = 0.015) for spline 3. Adjustment for temperature and smear density per 500 WBC slightly strengthened the associations (OR 9.47 and 8.99, both p = 0.021).

Table 8. Logistic regression of thrombocytopenia vs. parasite density.

Logistic regression analysis of thrombocytopenia vs. parasite density. Significant associations were observed at lower and higher parasite densities (splines 1 and 3). CI, confidence interval.

The model demonstrated moderate discriminative ability with an AUC of 0.677 (95% CI: 0.622–0.732). This indicates that platelet counts are more sensitive to fluctuations in parasite density than hemoglobin or WBC counts.

4 Discussion

4.1 Overview of findings and incremental insight

This study provides an integrated, cross-sectional evaluation of how parasite density relates to clinical symptom burden and routine hematologic indices in children with Plasmodium falciparum malaria at first clinical contact in an endemic setting. While several individual associations examined here have been reported previously, the incremental contribution of this study lies in the comparative and integrative interpretation of these relationships within a single pediatric cohort using a consistent analytical framework.

Our findings demonstrate that clinical symptom burden and severity exhibit a non-linear, density-dependent distribution at presentation, with more symptomatic phenotypes clustering among children presenting with higher parasite densities. In contrast, thrombocytopenia shows a strong and consistent association with parasite density across density ranges, suggesting greater sensitivity to contemporaneous parasitemia. Hemoglobin concentration and leukocyte counts, however, display weak or absent associations with parasite density at presentation, despite their frequent use in routine clinical assessment.

By examining clinical symptoms, platelet count, hemoglobin, and white blood cell count concurrently, our findings highlight important differences in how rapidly and sensitively different physiological systems reflect contemporaneous parasite burden. In particular, platelet counts appear to respond more dynamically to acute parasitemia than hemoglobin or leukocyte measures, which may reflect cumulative or heterogeneous processes rather than real-time infection intensity.

Thus, the principal insight offered by this study is not the identification of novel malaria manifestations, but rather a clearer delineation of which routinely available clinical and laboratory markers are most informative of parasite density at the point of presentation, and which are not. This distinction has practical implications for triage and early clinical interpretation in settings where quantitative parasite measurement may be delayed or unavailable.

4.2 Parasite density and symptomatology: cross-sectional density–symptom relationships

Our findings demonstrate a clear cross-sectional association between parasite density and both total symptom count and symptom severity index at presentation, consistent with prior studies reporting non-linear parasite–host relationships in malaria (Buffet et al., 2011; Ashley et al., 2018; Hopkins et al., 2020). These patterns likely reflect the combined influence of parasite biomass and contemporaneous host inflammatory responses at the time of clinical assessment.

At lower to moderate levels of parasitemia, host immune mechanisms may limit overt clinical expression despite active infection. In contrast, children presenting with higher parasite densities were more likely to exhibit greater symptom burden and severity, plausibly reflecting intensified innate immune activation and cytokine release (e.g., TNF-α, IL-6) at higher parasite loads (Chotivanich et al., 2000; Clark et al., 2006). Importantly, these associations describe variation in clinical presentation across parasite density ranges measured at a single time point, rather than symptom evolution or disease progression over time.

Gastrointestinal and constitutional symptoms—including headache, vomiting, abdominal pain, anorexia, and fatigue—were the predominant manifestations observed at presentation. This symptom profile is characteristic of uncomplicated pediatric malaria and reflects systemic inflammatory and pyrogenic responses rather than localized organ-specific pathology (Cowman et al., 2016; Folarin et al., 2021). Similar patterns have been reported in pediatric cohort studies from Kenya, Ghana, and Thailand, where symptom burden increased non-linearly with rising parasite density, and higher parasitemia was associated with greater frequency of constitutional and gastrointestinal symptoms, despite substantial inter-individual variability related to age and transmission intensity (Dondorp et al., 2004; Dondorp and Day, 2007; Dondorp et al., 2011; Cunnington et al., 2013).

4.3 Anemia severity and parasite density: a multifactorial relationship

Despite malaria’s well-established role as a leading cause of anemia, our analysis found no significant direct relationship between parasite density and anemia severity at presentation. This is biologically plausible for several reasons. First, malaria-induced anemia is a cumulative phenomenon, often peaking days to weeks after initial parasitemia, when hemolysis of both parasitized and non-parasitized red blood cells occurs (Fairhurst and Wellems, 2010; Eckhoff, 2011). Second, dyserythropoiesis and bone marrow suppression, induced by inflammatory cytokines, play an important role and do not always correlate with parasite density at a single time point (Gething et al., 2011; Grigg et al., 2018).

Third, in endemic settings, anemia is frequently multifactorial — nutritional deficiencies, chronic inflammation, hemoglobinopathies (e.g., sickle trait, α-thalassemia), helminth infections, and repeated subclinical parasitemia all contribute (Haldar and Mohandas, 2009; Idro et al., 2010; Hendriksen et al., 2012; Hanson et al., 2015). Indeed, several African studies have shown that while parasite prevalence correlates with anemia burden at the population level, parasite density at individual presentation often explains only a small fraction of hemoglobin variability (John et al., 2006; Jenkins and Price, 2010).

This finding has major clinical implications: hemoglobin alone is not a reliable proxy for parasite burden, and severe anemia may not coincide temporally with peak parasitemia. Monitoring hemoglobin trajectories after parasite clearance is essential in clinical care, especially for high-risk groups such as children under five.

4.4 Thrombocytopenia as a sensitive indicator of contemporary parasite burden

Consistent with extensive prior literature, thrombocytopenia emerged as the hematologic parameter most strongly associated with parasite density in this cohort, particularly across lower and higher parasite density ranges (Maitland et al., 2011; Kassebaum et al., 2014; MalariaGEN (Malaria Genomic Epidemiology Network), 2014; Maitland, 2015). Rather than redefining thrombocytopenia as a novel feature of malaria, our findings refine its interpretation by demonstrating its relative sensitivity to contemporaneous parasite burden compared with other routinely measured hematologic indices.

Several biological mechanisms likely underpin this association. Acute malaria is known to induce splenic sequestration of platelets, immune-mediated platelet destruction, and platelet activation with subsequent consumption during interactions with infected erythrocytes (Miller et al., 2002; Newton et al., 2003; Ochola-Oyier et al., 2010). In addition, experimental studies suggest that platelets may directly interact with and contribute to the killing of intraerythrocytic parasites, a process that may further accelerate platelet depletion during active infection (Pain et al., 2001). Together, these mechanisms provide biological plausibility for the observed density-dependent relationship and help explain why platelet counts appear to reflect acute parasitemia more dynamically than hemoglobin or leukocyte measures.

Importantly, platelet counts are known to decline early during infection and to recover rapidly following parasite clearance, often preceding hemoglobin recovery (Perkins and Were, 2010). In this context, our analysis shows that, at presentation, platelet count carries more immediate information about parasite density than hemoglobin or leukocyte count. The moderate discriminative performance of the thrombocytopenia model (AUC = 0.677) supports its potential role as a pragmatic, adjunctive indicator of parasite burden, rather than as a standalone diagnostic or prognostic marker.

While previous meta-analyses have documented thrombocytopenia in up to 80% of malaria cases (Perkins et al., 2011; Phyo et al., 2016), and some studies have linked its severity to adverse outcomes (Roca-Feltrer et al., 2008b; Rogerson and Carter, 2008), our findings contribute by situating thrombocytopenia within a broader comparative framework. Specifically, we show that among commonly available laboratory parameters, platelet count most consistently reflects parasite density at first clinical contact, whereas anemia severity and leukocyte counts appear less informative in this regard.

This distinction is clinically relevant in resource-limited settings, where clinicians frequently rely on a combination of symptoms and basic laboratory tests to infer disease burden. Our results support the use of platelet count—interpreted alongside symptom burden—as a useful contextual marker of acute parasitemia, while reinforcing the limitations of hemoglobin and WBC as real-time indicators of parasite load.

4.5 Weak associations of WBC count and parasite density

Our study found no meaningful association between parasite density and total WBC count, which is expected given the complexity of leukocyte dynamics in malaria. Total leukocyte counts often remain within the reference range in uncomplicated malaria (Safeukui et al., 2008; Rowe et al., 2009). In some patients, mild leukopenia occurs due to splenic pooling of lymphocytes and neutrophils (Seydel et al., 2006), while others may exhibit relative neutrophilia or monocytosis depending on disease stage and immune activation (Siqueira et al., 2014).

Because WBC responses are modulated by host immune history, co-infections, and age, their utility as a surrogate marker for malaria infection is limited. Similar observations have been reported in large cohort studies in Kenya, Ghana, and Thailand (Tangpukdee et al., 2009; Smith et al., 2013; Teo et al., 2016).

4.6 Integrating clinical and hematologic signals for risk stratification

Our results highlight the complementary value of clinical and hematologic indicators in malaria case management. Symptom burden appears to reflect the acute inflammatory response and underlying parasite replication dynamics at presentation, providing an immediate clinical signal of disease activity. Platelet counts, in contrast, function as a sensitive early hematologic marker of parasitemia, demonstrating stronger density-dependent variation than other routinely measured laboratory indices. Hemoglobin levels seem to capture longer-term or cumulative effects of infection rather than contemporaneous parasite load, while total white blood cell counts remain largely uninformative in uncomplicated cases due to their variability and modulation by host immune factors.

This layered interpretation aligns with recent efforts to integrate simple laboratory measures with clinical features for more accurate risk stratification in endemic settings (Tilley et al., 2011; Turner et al., 2013). For example, the combination of fever + thrombocytopenia has been shown to improve diagnostic specificity for malaria over fever alone (Van Wolfswinkel et al., 2013).

4.7 Biological plausibility: non-linear host–parasite relationships at presentation

A key methodological feature of this study was the use of natural cubic splines to model non-linear associations between parasite density and clinical as well as hematologic manifestations at presentation. Malaria pathophysiology involves complex, non-linear host–parasite interactions, in which inflammatory responses, endothelial activation, and hematologic perturbations do not increase proportionally with parasite burden (Wassmer et al., 2003; White et al., 2013).

Rather than implying temporal “thresholds” or transition points, the spline-derived inflection patterns observed in this study represent parasite density ranges at which particular clinical or laboratory features are more frequently observed at the time of assessment. Traditional linear models may obscure such cross-sectional heterogeneity by averaging effects across the entire density spectrum. In contrast, spline modeling allows risk or symptom burden to vary flexibly across parasite densities without assuming monotonic or proportional relationships.

Similar non-linear modeling approaches—including spline regression, generalized additive models, and density-threshold frameworks—have been applied in longitudinal cohort studies to examine parasite density–fever relationships and the development of immune tolerance across endemic settings in Africa and Asia (Wilson et al., 2010; White et al., 2014; World Health Organization, 2022). Although those studies focused on temporal dynamics and repeated measurements, our cross-sectional analysis demonstrates that comparable non-linear patterns are already evident at first clinical contact. This finding supports the biological plausibility of density-dependent variation in host responses observable at presentation.

Overall, the application of spline-based models in the present study provides a descriptive framework for characterizing how clinical and hematologic manifestations are distributed across the parasite density spectrum at enrollment, without implying disease progression, prognostic thresholds, or causal transitions.

4.8 Implications for clinical practice and malaria control

From a clinical standpoint, the findings underscore the utility of platelet counts as a triage tool and symptom clusters as early warning signals for moderate-to-high parasitemia. This may be particularly relevant in primary health facilities with limited access to parasitological confirmation.

From a public health perspective, recognizing non-linear symptom-parasite relationships can help refine case definitions in surveillance systems, especially in low-to-moderate transmission areas where clinical diagnosis remains common. Additionally, these data can inform the design of risk stratification algorithms integrating simple clinical and laboratory parameters to prioritize treatment and referral decisions.

4.9 Strengths and limitations

This study has several notable strengths. First, the use of flexible spline-based models allowed us to characterize non-linear, cross-sectional relationships between parasite density and clinical as well as hematologic manifestations at presentation, without imposing restrictive linear assumptions. Second, the integration of detailed symptom profiling with routine hematologic indices enabled a comparative evaluation of how different physiological systems reflect contemporaneous parasite burden. Third, the study population represents children presenting with uncomplicated malaria at a primary-care level in a high-transmission setting, enhancing the relevance of the findings to similar endemic contexts.

Several limitations warrant consideration. The cross-sectional design restricts inference on temporal dynamics, including symptom evolution and delayed hematologic effects such as post-infectious anemia. In addition, important contributors to anemia—such as nutritional status, hemoglobinopathies, and chronic or co-infections—were not measured and may have attenuated observed associations with parasite density. Platelet and leukocyte subpopulations (e.g., neutrophil-to-lymphocyte ratio or platelet indices) were also not examined and could provide further mechanistic insight.

Some regression models, particularly those evaluating rare outcomes such as severe symptoms (grade ≥ 3) and hematologic grade ≥ 3 events, yielded wide confidence intervals. These reflect data sparsity and low event counts at extreme parasite density ranges, rather than model misspecification. To evaluate the robustness of these findings, we conducted multiple sensitivity analyses, including varying spline degrees of freedom, excluding extreme parasite density values, and refitting sparse models using bias-reduced estimation. Across these analyses, point estimates remained directionally consistent, although precision remained limited for rare outcomes.

We considered simpler linear modeling strategies; however, such approaches tended to obscure density-dependent variation observed at higher parasite burdens by averaging effects across the full density spectrum. Given the biological plausibility of non-linear host–parasite interactions, spline models were retained to provide a more faithful descriptive representation of the data. Nevertheless, estimates derived from sparse outcome strata should be interpreted cautiously and are best viewed as exploratory rather than definitive.

Finally, this was a single-site study, and external validation in independent cohorts and longitudinal designs will be important to confirm the generalizability and stability of these findings.

4.10 Future directions

Future research should focus on longitudinal cohort studies with serial hematologic measurements, to capture the temporal evolution of anemia and thrombocytopenia during and after infection. Integrating immune biomarkers, such as cytokines and hepcidin, could clarify mechanisms driving differential hematologic responses (Yeo et al., 2008; Zimmemann et al., 2018; Zuzarte-Luis and Mota, 2018).

Multicenter studies would enable external validation and help refine parasite density thresholds for clinical decision-making. Incorporating platelet counts and symptom clusters into point-of-care diagnostic algorithms may enhance early case detection in resource-limited settings.

5 Conclusion

This study provides important insights into the complex, non-linear relationship between parasite density and clinical as well as hematologic manifestations of acute Plasmodium falciparum malaria in children presenting at a primary care level in Ilorin, Nigeria. By integrating standardized symptom profiling with robust laboratory assessment and flexible statistical modeling, we demonstrated that higher parasite densities are associated with increased symptom burden and specific hematologic alterations, including thrombocytopenia and progressive anemia.

These findings highlight the clinical utility of parasite density as an early risk stratification marker and support its potential role in guiding triage, monitoring disease severity, and improving case management in endemic settings. Strengthening routine clinical and laboratory surveillance at the point of care can enhance early recognition of severe or complicated malaria, ultimately contributing to better outcomes.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Formatting of the references.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmala.2026.1751312/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Isaac Kweku Quaye, Regent University College of Science and Technology, Ghana

Reviewed by: Adilson José DePINA, CCS-SIDA/MoH, Cabo Verde

Sompob Saralamba, Mahidol Oxford Tropical Medicine Research Unit (MORU), Thailand

Simon Kojo Appiah, Howard University, United States