Association between ovalocytosis and Plasmodium infection: a systematic review and meta-analysis – Scientific Reports

Search results

A total of 905 articles were retrieved from our database searches: MEDLINE (n = 88), Embase (n = 160), Scopus (n = 172), PubMed (n = 114), Ovid (n = 295), and ProQuest (n = 76). After removing duplicate records (n = 311), the titles and abstracts of the remaining records were screened (n = 594) to exclude non-relevant records (n = 443). Then, the remaining records (n = 151) were assessed for full-texts and eligibility. After applying the exclusion criteria, 133 records were excluded with specific reasons: reviews (n = 68), in vitro studies (n = 30), case reports (n = 10), gene mutation or polymorphism (n = 8), no abstract available (n = 4), conference abstracts (n = 2), books (n = 2), no information on ovalocytosis (n = 2), letters or news article (n = 2), inability to extract data (n = 2), unavailability of full text (n = 1), guideline (n = 1) and duplicate article (n = 1). The searches in Google Scholar identified 876 articles, none of which met the eligibility criteria. Finally, 16 studies^{15,16,17,26,27,28,29,30,31,32,33,34,35,36,37,38} were included for qualitative and quantitative data synthesis (Fig. 1).

General characteristics of the included studies

The included studies were published between 1977 and 2015, with the majority (43.8%) published between 2000 and 2010^{16,17,28,31,32,35,37}. Study designs included cross-sectional studies (50.0%)^{16,17,28,30,31,34,36,37}, case–control studies (18.8%)^26,27,29, cohort studies (25.0%)^32,33,35,38 and both cohort and case–control studies (6.3%)¹⁵. Most studies were conducted in Papua New Guinea (68.8%)^{15,16,26,28,29,32,33,34,35,36,38}, with the remainder in Indonesia (18.8%)^17,31,37, Malaysia (6.3%)³⁰ and Thailand (6.3%)²⁷. Most studies enrolled patients infected with single or mixed infections of P. falciparum and P. vivax (56.3%)^{16,17,27,28,29,31,33,34,37}, single or mixed infections of P. falciparum, P. vivax and P. malariae (18.8%)^15,30,36, and infection with P. falciparum only (12.5%)^26,35. The majority of study participants were children (50.0%)^{15,16,26,32,33,34,36,37}, followed by pregnant women (18.8%)^28,35,38. Most of the included studies tested for SAO using a polymerase chain reaction (PCR) assay to identify band 3 deletions (75.0%)^{15,16,17,26,28,31,32,33,34,35,37,38}, and the remainder investigated ovalocytosis using microscopy (25.0%)^27,29,30,36. Malaria parasites were mainly detected using microscopy (68.8%)^{16,17,26,27,28,29,30,31,34,35,36}, followed by a combination of microscopy with PCR (18.8%)^15,32,38, a combination of microscopy with RDT (6.3%)³³ or PCR only (6.3%)³⁷ (Table 1). Table S2 presents details of the included studies.

Risk of bias

All three case–control studies^26,27,29 and three of the cohort studies^15,32,33 scored the maximum of 9/9 stars. Two studies^35,38 scored 8/9 stars: one did not demonstrate the outcome of interest at the start of the study³⁵ and the other lacked selection of the non-exposed cohort³⁸. ‘The NOS scale for case–control studies was adapted to assess the risk of bias in cross-sectional studies. Six cross-sectional studies scored 8/8 stars^{16,28,30,31,34,36}, and two^17,37 scored 7/8 stars because they lacked a ‘definition of controls’ (Table S3).

Qualitative synthesis

Seven studies (43.8%) showed that ovalocytosis might protect against malaria infection^15,27,29,30 or severe disease^15,26,28,36. Apibal et al.²⁷ reported that patients with malaria had an increased percentage of ovalocytes compared with uninfected patients (P. falciparum: mean, 6.3 cells; P. vivax: mean, 8.3 cells; normal individuals: mean, 0.6 cells). Benet et al.²⁸ reported that severe malaria infections in placental tissues were less common in pregnant women with SAO than in pregnant women in the control group. Cattani et al.²⁹ reported that patients with ovalocytosis had a lower infection rate of P. falciparum (P = 0.044), P. vivax (P = 0.009) and all species of malaria combined (P = 0.013) compared with those without ovalocytosis. Foo et al.³⁰ demonstrated fewer ever-positive ovalocytes for P. falciparum (P < 0.05) or any parasite species (P < 0.05) compared with the controls. Additionally, malaria parasitaemia was lower in individuals with ovalocytes compared with the controls. Rosanas-Urgell et al.¹⁵ reported that SAO was associated with a statistically significant reduction of 46% in the incidence of clinical P. vivax episodes and a 52% reduction in P. vivax blood-stage reinfection. Furthermore, in the case–control study, SAO was associated with protection against severe P. falciparum malaria (OR = 0.38, P = 0.014) but not with protection against uncomplicated P. falciparum malaria. Allen et al.²⁶ reported the absence of SAO band 3 in 68/68 (100%) of children with cerebral malaria compared with 6/68 (8.8%) matched community controls (OR = 0, 95% CI 0.0–0.85). Serjeantson et al.³⁶ suggested that patients with ovalocytosis were more resistant to severe malaria than those with normocytes because the ratio of parasitaemia in 112 children with ovalocytes compared with 741 children with normocytes was 1.05 for P. falciparum; 0.90 for P. vivax; 0.54 for P. malariae, and 0.91 for infection with any Plasmodium species.

Eight studies (8/16, 50%) showed no association of SAO with malaria infection or severity^{16,17,31,32,34,35,37,38}. Fowkes et al.¹⁶ found no association of SAO with P. falciparum prevalence (P = 0.29) or with mean parasite density (P = 0.66). Kimura et al.¹⁷ found no difference in the prevalence of SAO between patients with malaria and controls (P > 0.8) and no difference in the frequency of SAO between patients with P. falciparum and P. vivax malaria. Although Kimura et al.³¹ discovered a link between a higher rate of ovalocytes and a lower risk of malaria infection, they found that SAO did not protect against malaria infection. Lin et al.³² showed no significant associations between P. falciparum infection and SAO (P > 0.2). O’Donnell et al.³⁴ demonstrated no difference in the proportion of SAO in acute malaria and the community controls (8.8% vs 6.6%, P = 0.57). However, the degree of ovalocytosis was significantly lower in children with SAO during acute malaria compared with the community controls (P = 0.025). O’Donnell et al.³⁵ showed that the SAO genotype was not associated with the frequency of placental P. falciparum infection (placental parasitaemia with a normal genotype: 24.5%, placental parasitaemia with SAO: 21.1%). Shimizu et al.³⁷ reported no significant difference in the prevalence of asymptomatic malaria infection between participants with or without SAO (P > 0.05), and Stanisic et al.³⁸ reported no significant difference in the frequency of malaria infection among participants with or without SAO (P = 0.62). Manning et al.³³ reported a frequently higher proportion of ovalocytosis in P. falciparum (5.7%) than P. vivax (3.8%) and mixed infections (4.3%).

Risk of malaria among participants with ovalocytosis

Eleven studies^{15,16,17,26,29,30,34,35,36,37,38} were included in the meta-analysis of the risk of malaria among participants with ovalocytosis. Overall, the results demonstrated no difference in the log OR of malaria between patients with and without ovalocytosis (P = 0.81, log OR = 0.06, 95% CI − 0.44 to 0.56, I²: 86.20%; 11 studies; Fig. 2). The meta-regression analyses using study design, country, participants’ group, Plasmodium species, method used to investigate ovalocytosis and method for malaria detection as covariates showed that none of these impacted the pooled effect estimate. Thus, the subgroup analysis was discontinued (Table S4). Since Rosanas-Urgell et al.¹⁵ showed that SAO was associated with a statistically significant reduction in the incidence of P. vivax infection, the meta-analysis was refined to P. vivax exclusively to prevent any interference from a concomitant P. falciparum infection. However, the results showed no difference in the log OR of P. vivax infection between patients with and without ovalocytosis (P = 0.83, log OR = 0.07, 95% CI: − 0.61 to 0.75, I²: 65.39%; 5 studies; Fig. 3).

Forest plot showing the pooled log odds ratio (OR) of the association between Southeast Asian ovalocytosis (SAO) and malaria infection. Abbreviations: Malaria (O) malaria infection in SAO RBCs; Malaria (non-O) malaria infection in non-SAO RBCs; CI confidence interval; blue square effect estimate (log OR); crimson diamond pooled log OR in each subgroup; green diamond pooled log OR in all included studies.

Forest plot showing the pooled log odds ratio (OR) of the association between Southeast Asian ovalocytosis (SAO) and P. vivax infection. Abbreviations: P. vivax (O), P. vivax infection in SAO RBCs; P. vivax (non-O), P. vivax infection in non-SAO RBCs; CI confidence interval; blue square effect estimate (log OR); green diamond, pooled log OR in all included studies.

Sensitivity analysis

The leave-one-out sensitivity analysis did not identify any outliers in the meta-analysis (P values in re-run analyses > 0.05; Fig. 4), indicating the robustness of the results. Meta-analysis using fixed-effect models was performed to test whether the difference in statistical models affected the pooled effect estimate. The results showed a decreased log OR of malaria among patients with ovalocytosis compared with those without ovalocytosis (P < 0.01, log OR = 0.82, 95% CI 0.49–1.14, I²: 95.10%; 3 studies; Fig. 5).

Leave-one-out sensitivity analysis to identify outliers in the meta-analysis of the log odds ratio between Southeast Asian ovalocytosis and malaria infection. Abbreviations: CI confidence interval; green dot effect estimate.

Sensitivity analysis using the fixed-effects model for the pooled log odds ratio between Southeast Asian ovalocytosis and malaria infection. Abbreviations: CI confidence interval; blue square effect estimate (log OR); crimson diamond pooled log OR in each subgroup; green diamond pooled log OR in all included studies.

Publication bias

Publication bias was assessed using a funnel plot and Egger’s test. The funnel plot showed the asymmetrical distribution of the log OR and indicated the standard error of the log OR (Fig. 6). Egger’s test demonstrated that the small-study effect was not significant (P = 0.68). The results of both tests indicated a publication bias among the included studies. Although the trim-and-fill method was applied to correct for publication bias, the subsequent results showed no difference in the log OR of malaria among patients with ovalocytosis compared with those without ovalocytosis (log OR = 0.297, 95% CI − 0.126 to 0.721).

Funnel plot demonstrating the asymmetrical distribution of the log odds ratio of individual studies.