C-reactive protein as an early biomarker for malaria infection and monitoring of malaria severity: a meta-analysis

This study investigated whether C-reactive protein (CRP) can be used as a marker for the early detection and monitoring of malaria severity. Potentially relevant studies were searched in Medline (PubMed), Scopus, and Web of Science. Differences in CRP between (1) severe malaria and uncomplicated malaria, (2) uncomplicated malaria and asymptomatic malaria, (3) uncomplicated malaria and febrile/healthy controls, and (4) asymptomatic malaria and febrile/healthy controls were estimated using random-effects models. Twenty-nine studies were included for meta-analysis. The results of meta-analysis demonstrated higher mean CRP levels in (1) patients with severe malaria compared with uncomplicated malaria (p < 0.001, standard mean difference [SMD]: 1.52, 95% confidence interval [CI]: 0.91–2.12, I2: 95.1%), (2) patients with uncomplicated malaria than in those with asymptomatic malaria (p: 0.001, SMD: 1.65, 95% CI: 0.67–2.62, I2: 96.7%), (3) patients with uncomplicated malaria compared with febrile/healthy controls (p < 0.001, SMD: 2.38, 95% CI: 1.37–3.40, I2: 98.5%), and (4) patients with asymptomatic malaria compared with febrile/healthy controls (p < 0.001, SMD: 2.55, 95% CI: 1.60–3.50, I2: 99.2%). This study demonstrated CRP levels are a biomarker for the early detection and monitoring of malaria severity.

Data synthesis. In the meta-analysis, the mean and standard deviation (SD) of CRP between two groups were used to estimate the pooled standard mean difference (SMD) and 95% confidence interval (CI). The SMD and 95% CI of each outcome were estimated using the random-effects models provided in the STATA Statistical Software version 15.0 (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC). If the median and range of CRP were reported by the included studies, the mean and SD were calculated from the median and range as described previously 24 . Additionally, if the included studies reported separate mean and SD for the clinical symptoms of malaria including CRP in cerebral and severe anemia, the mean and SD were combined as a single "severe malaria" as described previously 25 . The heterogeneity among the included studies was assessed using the Cochrane Q and I 2 statistics. Cochrane Q statistic with a p-value of less than 0.05 indicated a significant heterogeneity among the included studies, while I 2 statistics of more than 50% indicated a substantial heterogeneity among the included studies. When heterogeneity was significant or substantial, the random-effects model was applied to the pooled analysis. Conversely, the fixed-effects model was applied to the pooled analysis if nonsignificant heterogeneity was demonstrated. Meta-regression analysis was performed to assess potentially important covariates that might have a substantial impact on the heterogeneity. These covariates included the mean age of participants, male percentage, continents (Asia, Africa, America, Europe), types of infection (Plasmodium spp.), and types of control (febrile or healthy controls). Then, subgroup analyses were conducted for these covariates. A sensitivity test was performed to assess (1) the pooled SMD after excluding studies that reported the median CRP, (2) the pooled SMD after excluding studies with six stars' quality, and (3) the pooled SMD using a fixed-effects model compared with the random-effects model. The publication bias among the included studies was assessed by visualization of the funnel plot and Egger's test for asymmetry. If the Egger's test showed asymmetry, the contour-enhanced funnel plot was further assessed to determine whether the asymmetry was due to publication bias or other factors.
Quality of the included studies. The quality of the included studies was assessed using the NOS, which has a maximum total of 7 stars. Sixteen studies were rated 7 stars in total, while 14 studies were rated 6 stars in total as febrile patients had been used as a control group (Table S3) (Fig. 3).
Subgroup analysis of types of infection demonstrated a higher mean CRP level in patients with severe malaria than in uncomplicated malaria patients in studies of patients with P. falciparum (p < 0.001, SMD: 1.19, 95% CI: The difference in the mean CRP level between patients with uncomplicated and asymptomatic malaria. The difference in the mean CRP level between patients with uncomplicated and asymptomatic malaria was estimated from four studies 27,33,40,46 . The results of the individual study demonstrated a higher mean CRP level in patients with uncomplicated malaria than in those with asymptomatic malaria among the three studies 33,40,46 , while a study by Andrade et al. 27 showed no difference in the mean CRP level between the two groups. The highest difference in the mean CRP level was reported in a study by Jakobsen et al. 40 . The pooled analysis of the four studies showed a higher mean CRP level in patients with uncomplicated malaria than in those with asymptomatic malaria (p < 0.001, SMD: The difference in the mean CRP level between patients with uncomplicated malaria and febrile/ healthy controls. The difference in the mean CRP level between patients with uncomplicated malaria and febrile/healthy controls was estimated from 12 studies 26,27,30,36,37,39,40,[44][45][46]50,54 . A meta-analysis of the difference in mean CRP level between patients with uncomplicated malaria and febrile/healthy controls were divided into three subgroups: uncomplicated malaria and febrile controls, uncomplicated malaria and healthy controls, and uncomplicated malaria and febrile/healthy controls because previous studies 27, 46 reported the mean/median CRP levels of febrile and healthy controls.   www.nature.com/scientificreports/ In the healthy control subgroup, the results of the individual study demonstrated a higher mean CRP level in patients with uncomplicated malaria than in healthy controls among the four studies 26,30,40,50 . The pooled analysis in these subgroups demonstrated a higher mean CRP level in patients with uncomplicated malaria than in healthy controls (p < 0.001, SMD: 3.80, 95% CI: 2.78-4.83, I 2 : 86%, 4 studies) (Fig. 6).
In the healthy and febrile control subgroup, the results of the individual study demonstrated a higher mean CRP level in patients with uncomplicated malaria than in febrile/healthy controls in a study by Mendonça et al. 46 , while no difference was shown in a study by Andrade et al. 27 . The pooled analysis in these subgroups demonstrated no difference in the mean CRP level in patients with uncomplicated malaria than in febrile/healthy controls (p: 0.172, SMD: 1.32, 95% CI: − 0.57-3.22, I 2 : 98.7%, 2 studies).
In the febrile subgroup, the results of the individual study demonstrated a higher mean CRP level in patients with uncomplicated malaria than in febrile controls among five studies 37,39,44,45,54 , while a lower mean CRP level in patients with uncomplicated malaria than in febrile controls was demonstrated in a study by Eriksson et al. 36 . The pooled analysis in these subgroups demonstrated a higher mean CRP level in patients with uncomplicated malaria than in febrile controls (p: 0.028, SMD: 1.80, 95% CI: 0.19-3.4, I 2 : 98.9%, 6 studies).
Meta-regression analysis using the mean age of participants, male percentage of participants, Plasmodium spp., types of control, or continents as covariates demonstrated no substantial impact of the mean age of participants Subgroup analysis of continents demonstrated a higher mean CRP level in patients with asymptomatic malaria than in febrile/healthy controls in studies conducted in Africa (p < 0.001, SMD: 3.39, 95% CI: 1.93-4.85, I 2 : 99.2%, seven studies) whereas no difference in the mean CRP level was observed in patients with asymptomatic malaria compared with febrile/healthy controls in studies conducted in America (p: 0.76, SMD: 3.22, 95% CI: − 0.05-1.56, I 2 : 93.7%, two studies) (Fig. 10).
Subgroup analysis of types of infection demonstrated a higher mean CRP level in patients with asymptomatic malaria than in febrile/healthy controls in studies of patients with P. falciparum (p < 0.001, SMD: 3.39, 95% CI: 1.93-4.85, I 2 : 99.2%, seven studies). No difference in the mean CRP level was observed in patients with asymptomatic malaria compared with febrile/healthy controls in studies of P. vivax (p: 0.064, SMD: 0.76, 95% CI: − 0.05-1.56, I 2 : 93.7%, two studies) (Fig. 11).   Differences in CRP levels between patients with severe and uncomplicated malaria were estimated after studies reporting the mean CRP were excluded. Results demonstrated a higher mean CRP level in patients with severe malaria than in those with uncomplicated malaria (p: 0.006, SMD: 1.21, 95% CI: 0.35-2.06, I 2 : 96%, six studies) (Supplementary Fig. 1).
2. Studies with six stars' quality were excluded. Differences in CRP levels between patients with severe and uncomplicated malaria were estimated after studies with six stars' quality were excluded. Results demonstrated a higher mean CRP level in patients with severe malaria than in those with uncomplicated malaria (p: 0.001, SMD: 1.72, 95% CI: 0.41-3.04, I 2 : 96.3%, five studies) (Supplementary Fig. 2).

Uncomplicated malaria and asymptomatic malaria.
Using the effect size from four studies, the funnel plot demonstrated an asymmetrical distribution indicating publication bias among the included studies (Fig. 13). Egger's test demonstrated no significant small-study effects (p: 0.94) among the four studies, indicating no publication bias. Therefore, a trim-and-fill analysis was not conducted.
3. Uncomplicated malaria and febrile/healthy controls. Using the effect size from 12 studies, the funnel plot demonstrated an asymmetrical distribution indicating publication bias among the included studies. Egger's test demonstrated significant small-study effects (p: 0.012) among the 12 studies (Fig. 14), indicating publication bias. A trim-and-fill analysis was conducted. Results showed a higher mean CRP level in patients with uncomplicated malaria than in those with febrile/healthy controls as estimated by the random-effects model (p: 0.04, SMD: 1.02, 95% CI: − 0.031-2.00, 17 studies) and by the fixed-effects model (p < 0.001, SMD: 0.81, 95% CI: 0.7-0.92, 17 studies). 4. Asymptomatic malaria and febrile/healthy controls. Using the effect size from 10 studies, the funnel plot demonstrated an asymmetrical distribution indicating publication bias among the included studies. Egger's test demonstrated no significant small-study effects (p: 0.152) among the 10 studies, indicating no publication bias (Fig. 15). A trim-and-fill analysis was not conducted.

Discussion
In this study, our meta-analysis demonstrated that CRP levels were significantly higher in patients with severe malaria than in those with uncomplicated malaria, in patients with uncomplicated malaria than in those with asymptomatic malaria, in patients with uncomplicated malaria than in febrile or healthy controls, and in patients with asymptomatic malaria than in febrile or healthy controls. These results suggest CRP might be used as an early biomarker for malaria infection and the monitoring of malaria severity. When using CRP level as a marker of malaria, a previous study showed that a cut-off value for CRP level of 10.8 mg/L could discriminate malaria from healthy controls 55 . Nevertheless, another study conducted in Southern Uganda demonstrated that about half of patients (58%) with malaria had elevated serum CRP levels greater than 50 µg/L, whereas 62% of patients without malaria had elevated CRP levels above 50 µg/L, indicating a poor cut-off CRP level for discriminating malaria from non-malaria 56 . A previous study conducted in Ghana reported low positive predictive values of elevated CRP for malaria (32%) and suggested that CRP was not useful for predicting parasitemia and malaria 57 . Another study demonstrated an increased CRP level during malaria infection but it had a low specificity to differentiate malaria from septicemia 57 . Furthermore, increased CRP levels were found in bacterial infections 14,57,58 . Although increased CRP had a low positive predictive value for malaria, a previous study demonstrated the high negative predictive value of elevated CRP for malaria (97.0%), indicating the likelihood of no malarial parasites in the blood of patients with normal CRP level, allowing malaria   www.nature.com/scientificreports/ to be ruled out in febrile patients 57 . Therefore, the potential usefulness of CRP includes ruling out other febrile illnesses in areas where malaria is endemic such as in sub-Saharan Africa. However, further studies are required to explain how inflammatory responses compared with non-malarial inflammatory responses during malaria lead to a distinct pattern. This information might help clinicians make decisions about the diagnosis of febrile conditions in highly malaria-endemic areas and to treat patients accordingly.
A previous study conducted in Asian countries including Cambodia, Laos, and Thailand suggested elevated CRP might be used to discriminate malaria from viral infections although it could not discriminate malaria from bacterial infections 16 . Another study suggested that CRP in combination with hematological parameters including thrombocytopenia and/or leukopenia could differentiate between malaria, dengue, and enteric fever 59 . In addition, a CRP level greater than 5 mg/L could discriminate malaria from dengue infection with a sensitivity of 95% but with a poor specificity of 35% 59 . Another study suggested that a low CRP level suggested dengue fever whereas a lower mean platelet count combined with an elevated CRP level was more indicative of malaria infection in areas where both malaria and dengue were endemic 60 .
When using CRP level as a marker of malaria severity, a previous study showed that a cut-off value for CRP level of 18.5 mg/L discriminated severe malaria from uncomplicated malaria with 71.4% sensitivity and 68.7% specificity 55 . Although a CRP level less than 20 mg/L was suggested to be a strong indicator of uncomplicated malaria 61 , increased CRP levels correlated strongly with malarial parasite density, suggesting it also correlated with malaria severity 62 . In addition, increased CRP levels were found in patients with multiple complications compared with those with a single complication, and a higher CRP level was observed in patients who died compared to those who survived 47 . Furthermore, an increased CRP level was associated with a low hemoglobin concentration and a longer hospital stay, indicating a poor outcome for patients with malaria 63 . A combination of increased CRP level and other routine laboratory parameters might improve the ability of CRP to differentiate between severe and uncomplicated malaria. For example, a combination of elevated CRP with an erythrocyte sedimentation rate (ESR) greater than 34.5 mm in the first hour of diagnosis helped identify patients with uncomplicated malaria who might subsequently develop severe malaria 64 . The results of the meta-analysis supported that the elevated levels of CRP may help in the prognosis of disease severity among patients infected with malaria. However, a lower CRP concentration in patients with severe and fatal malaria than in those with uncomplicated   www.nature.com/scientificreports/ malaria had been observed, suggesting the inability to control the inflammatory response to infection; this may be particularly important for protection against cerebral malaria 65 . The role of CRP in malaria that is linked to the severity of the disease is correlated with nitrox oxide (NO), which is a toxic substance against P. falciparum 66 . Additionally, increased NO could activate neurons and damage erythrocytes, which might contribute to severe anemia or cerebral malaria in patients with severe malaria 67 .
A previous study reported that, when using CRP level as a marker of asymptomatic malaria, prolonged exposure to Plasmodium infections among asymptomatic malaria patients resulted in a chronic inflammatory response although the median CRP value in asymptomatic malaria patients did not differ from that in the healthy population 63 . In addition, CRP levels in asymptomatic malaria patients did not vary with high density parasitemia (more than 10,000 parasites/µL) and remained less than 3.5 µg/mL compared with uncomplicated malaria patients in whom the median CRP concentration was 116.4 µg/mL, indicating that high parasitemia alone did not stimulate an acute-phase response 63 . These results indicated that CRP could be a marker of malaria infection in those participants who live in communities in endemic areas without signs or symptoms of malaria.
Besides the usefulness of CRP as an early marker for malarial infection and severity, it could be used as a prognostic marker for the efficacy of malaria treatment as CRP level was reported to be decreased under malaria treatment 68,69 . Therefore, CRP measurement may be useful for the physician to follow-up on the efficacy of treatment of malaria. However, the use of CRP as a marker of malaria infection needs to be interpreted in combination with other parameters, including procalcitonin 69 , haptoglobin 70 , serum hepcidin 71 , blood transaminases 66 , and blood count parameters 3,4,60,72,73 , to help increase the sensitivity of this marker. A high CRP level with other routine laboratory parameters could help differentiate patients with uncomplicated malaria from those with asymptomatic malaria.
In Africa, where malaria is endemic, CRP was reported to aid the diagnosis of neonatal sepsis [74][75][76] . In addition, a combination of CRP and procalcitonin improved the accuracy of the diagnosis of neonatal sepsis 77,78 . A study of 624 apparently healthy volunteers in Ghana reported low circulating CRP levels in the healthy Ghanaian population, indicating an adverse environmental condition in a malaria-endemic area 79 . Therefore, interpreting the differences in CRP levels among different ethnic groups living in the same country needs to be considered.   Figure 11. Forest plot demonstrates the difference in the mean CRP level between patients with asymptomatic malaria and healthy/febrile controls by types of infections. SMD, standard mean difference; CI, confidence interval. www.nature.com/scientificreports/ Differences in CRP, particularly hs-CRP, which can detect a low amount of CRP (1-10 mg/L) 80 , were welldescribed in a meta-analysis involving 18,585 participants of African ancestry that reported higher hs-CRP levels in black residents in the United States compared with Hispanics, South Asians, Caucasians, and East Asians 81 . Nevertheless, in Africa hs-CRP appears to be a good marker for the early detection of malaria in asymptomatic individuals. Further studies are required to investigate the performance of hs-CRP to discriminate individuals with asymptomatic malaria from healthy individuals in those areas. This study has several limitations. First, a high heterogeneity among the included studies was observed. Although meta-regression and subgroup analyses of age, male percentage, continents, types of infection, or type of controls were performed, the heterogeneity was high, although the source of heterogeneity could not be explored; therefore, the pooled analysis needs to be carefully interpreted. Second, the number of studies included in each analysis was limited because some relevant studies were excluded due to incomplete data for CRP level or clinical status of patients with malaria presented in the literature. Third, there was publication bias among the included studies that analyzed differences in the mean CRP level between patients with severe malaria and uncomplicated malaria, as well as between uncomplicated malaria and febrile/healthy controls; therefore, the

Conclusion
This study demonstrated the possibility of CRP as a biomarker of malaria infection and severity. Using CRP in combination with other routine laboratory parameters could serve as a biomarker for the early detection and monitoring of malaria severity. Further studies with large sample sizes are needed to determine the performance of CRP to help clinicians in highly malaria-endemic areas make appropriate decisions on malaria treatment.  License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.