Severe Malarial Thrombocytopenia: A Risk Factor for Mortality in Papua, Indonesia

Background The significance of thrombocytopenia to the morbidity and mortality of malaria is poorly defined. We compared the platelet counts and clinical correlates of patients with and those without malaria in southern Papua, Indonesia. Methods Data were collated on patients presenting to a referral hospital between April 2004 and December 2012. Results Platelet measurements were available in 215 479 patients (23.4%), 66 421 (30.8%) of whom had clinical malaria. Patients with Plasmodium falciparum monoinfection had the lowest platelet counts and greatest risk of severe thrombocytopenia (platelet count, <50 000 platelets/µL), compared with those without malaria (adjusted odds ratio [OR], 6.03; 95% confidence interval [CI], 5.77–6.30]). The corresponding risks were 5.4 (95% CI, 5.02–5.80) for mixed infections, 3.73 (95% CI, 3.51–3.97) for Plasmodium vivax infection, and 2.16 (95% CI, 1.78–2.63) for Plasmodium malariae infection (P < .001). In total, 1.3% of patients (2701 of 215 479) died. Patients with severe malarial anemia alone (hemoglobin level, <5 g/dL) had an adjusted OR for death of 4.93 (95% CI, 3.79–6.42), those with severe malarial thrombocytopenia alone had an adjusted OR of 2.77 (95% CI, 2.20–3.48), and those with both risk factors had an adjusted OR of 13.76 (95% CI, 10.22–18.54; P < .001). Conclusions Severe thrombocytopenia identifies both children and adults at increased risk of death from falciparum or vivax malaria, particularly in those with concurrent severe anemia.

responsible for an increasing proportion of infections [1]. Anemia is a common manifestation of Plasmodium infection and is responsible for substantial morbidity, as well as for direct and indirect mortality [2,3]. Thrombocytopenia is also a common feature of malaria due to all Plasmodium species [4][5][6], but in the absence of significant bleeding it is not regarded as a defining clinical manifestation of severe malaria [7]. In most cases, thrombocytopenia is not associated with bleeding and requires no treatment, with the platelet count rapidly returning to normal after successful treatment of the malarial episode. Pregnant women and infants appear to be at increased risk of thrombocytopenia, although the adverse consequences of this are unclear [8,9]. Severe thrombocytopenia (defined as a platelet count of <50 000 platelets/μL) is reported in both P. falciparum and Plasmodium vivax infections, and although it has been associated with bleeding [10,11] and disseminated intravascular coagulation [12][13][14], hemorrhagic manifestations are unusual. More recently, low platelet counts have been associated with mortality in patients with P. falciparum [15] and P. vivax infections [16,17]. However, other studies have not demonstrated an association between thrombocytopenia and significant clinical risk [18,19].
The present study is part of a prospective surveillance of clinical and laboratory data from Mitra Masyarakat Hospital in southern Papua, Indonesia, an area where 4 species of malaria are coendemic. This analysis was conducted to establish the comparative platelet counts of patients infected by the different Plasmodium species and to define the associated risks of morbidity and mortality.

Study Site
Mimika district lies in south-central Papua, the easternmost province of Indonesia. Its geography, climate, and demographic characteristics have been described elsewhere [3,20]. In brief, malaria transmission occurs throughout the year but is limited to the lowland areas. The average annual incidence of parasitemia is estimated to be 876 episodes per 1000 people, with most cases due to P. falciparum. The prevalence of asexual parasitemia in 2005 was estimated to be 7.5% for P. falciparum, 6.4% for P. vivax, 1.9% for mixed infection, and 0.6% for Plasmodium malariae [20]. Until November 2008, Rumah Sakit Mitra Masyarakat (RSMM) was the only referral hospital in the district, and since 2008 approximately 80% of patients with malaria attending an inpatient facility in the district have been treated there. RSMM has a capacity of 110 beds, with a high-dependency unit, a 24hour emergency department, and an outpatient department that reviews approximately 300 patients per day, 6 days per week. The age distribution of all patient presentations peaks in infancy, with a second peak among individuals aged in their late 20s [3], whereas the absolute number of patient presentations with malaria peaks during the second year of life. Vivax malaria is the dominant cause of malaria in patients <3 years of age in both the outpatient and inpatient setting [9], and thereafter, P. falciparum is the most common malaria parasite [3,20].

Laboratory and Data Collection Procedures
Hospital protocols recommend that all patients presenting to the outpatient department with a fever and that all inpatients, regardless of diagnosis, should have a blood film performed for detection of malaria parasites. Microbiological diagnosis of malaria is based on a thick blood film examination, with confirmatory thin blood films and rapid diagnostic tests for P. falciparum also performed in some cases. Microscopy quality control of the hospital laboratory suggests >90% accuracy [21].
On their first presentation to RSMM, every patient is assigned a unique hospital record number, and this is used to link all clinical and laboratory data from all presentations. Demographic and administrative information is recorded by hospital clerks, along with the diagnosis from the attending physician (classified according to the International Classification of Diseases) and any deaths. For the purposes of analyses, ethnicity was categorized as Highland Papuan, Lowland Papuan, or non-Papuan, based on location of the clans' village(s). Complete blood counts are ordered according to clinical indication and are generated by coulter counter (JT Coulter, Ramsey, Minnesota).

Data Merging and Statistical Analyses
All statistical analyses were done in Stata, version 12.1. Clinical and hematology data were merged using the unique hospital record number and date of presentation. If >1 laboratory measurement was available for a single presentation, the minimum platelet count was taken ( Figure 1). The primary outcome in this study was the mean number of platelets per microliter associated with infection for each Plasmodium species, compared with patients without malaria. Secondary measures included the risk of severe thrombocytopenia, the population attributable fraction (PAR) of severe thrombocytopenia associated with infection by the different Plasmodium species, and all-cause mortality. Thrombocytopenia was defined as severe if the platelet count was <50 000 platelets/μL (approximately the fifth percentile) and very severe if the count was <20 000 platelets/μL (approximately the first percentile; Supplementary Figure 1). Severe anemia was defined as a hemoglobin concentration of <5 g/dL. Continuous data were analyzed using linear regression, and binary data (such as severe thrombocytopenia and death) were analyzed using logistic regression. Since some patients appeared in the database multiple times, robust standard errors were calculated using the clustered sandwich estimator.
For the purposes of these analyses, mixed infections were defined as concomitant infection with any combination of Plasmodium species. Univariable and multivariable analyses were performed for each of the following variables: infecting Plasmodium species (negative, P. falciparum, P. vivax, P. malariae, P. ovale, or mixed species), sex, ethnicity (non-Papuan, Highland Papuan, or Lowland Papuan), age group (<1 year, 1 to <5 years, 5 to <15 years, and ≥15 years), year of presentation (2004-2012), department (outpatient vs inpatient), and number of presentations with malaria in the preceding 2 months. Fractional polynomials were used to define the nonlinear relationship between age and the mean platelet count and risk of severe thrombocytopenia [22], but to maintain the stability of these models the following patients were excluded: patients with platelet counts of <5000 or >1 000 000 platelets/μL (338 [0.16%]), infants <1 week of age (1073 [0.50%]), and adults >63 years of age (the 99th percentile; 2205 [1.02%]).
Adjusted PAFs of severe thrombocytopenia were calculated from multivariable logistic regression models, using the punaf module for Stata, which derives PAFs by means of the formulae provided in Greenland and Dreschler [23]. Because of very small numbers, data from patients with P. ovale infections (30 [0.01%]) are included in the baseline values and the univariable analyses but excluded from the multivariable analyses.

Ethics Approval
Ethics approval for this study was obtained from the Health Research Ethics Committees of the University of Gadjah Mada, Indonesia, and the Menzies School of Health Research, Darwin, Australia. Since data were gathered from routine hospital surveillance, informed consent was not requested from participants. However, all records were anonymized, to ensure patient confidentiality.

Availability of Platelet Data
Overall, 215 479 presentations (23.4%) could be matched with at least 1 platelet count measurement ( Figure 1). A greater
Overall, compared with patients with neither severe anemia nor thrombocytopenia, the adjusted ORs for death were 5.21 (95% CI, 4.53-5.98) among those with severe anemia alone,   (Table 3). This relationship was apparent in both children and adults with malaria and in P. falciparum, P. vivax, and mixed infections (Table 4).
In the absence of severe anemia, the greatest risk of mortality associated with severe thrombocytopenia was among patients without malaria (adjusted OR, 6.21 [95% CI, 5.37-7.20]; P < .001), whereas the risk among patients with malaria was 2.77 (95% CI, 2.20 to 3.480; P < .001), with no difference between infecting species (Table 4). The overall PAF of death associated with severe thrombocytopenia was 14.6% (95% CI, 13.1%-16.0%); the full multivariable model for mortality is presented in Table 3. There was no significant difference in the risk

DISCUSSION
In this very large hospital-based surveillance study, almost two thirds of patients with acute malaria had thrombocytopenia ( platelet count, <150 000 platelets/μL), with 13% of patients presenting with platelet counts of <50 000 platelets/μL. The greatest risk of severe thrombocytopenia was in patients infected with P. falciparum, either alone or mixed (OR, 5.4-6.1), accounting for >40% of observed cases. Severe thrombocytopenia was associated with a 2.4-fold greater risk of admission to hospital and a 4.7-fold increased risk of death, rising to 16-fold when both severe anemia and severe thrombocytopenia were present (Table 3). Similar relationships between the risk of death and severe thrombocytopenia were seen in both children and adults with malaria and in cases of P. falciparum and P. vivax infections. Malaria causes a variety of hematological insults arising from hemolysis, host inflammatory response, hematopoietic suppression, and splenic pooling [24,25]. Severe anemia is an important prognostic indicator of fatal outcome, particularly in young children [3,26]. While thrombocytopenia is also extremely common, its contribution to morbidity and mortality has been less clear. In patients with falciparum malaria, severe disease and mortality are increased with severe thrombocytopenia [15,27], and more recently this has also been observed in patients with severe vivax malaria [16]. Many previous studies examining predictors of malaria mortality have not included platelet counts [7]. Other studies showing no relationship between malarial thrombocytopenia and mortality have been smaller and may have been underpowered [28,29].
Previous studies have shown a consistent inverse correlation between parasitemia at presentation and the platelet count [17], but our study did not record the peripheral parasite count routinely, and hence we were unable to explore this. In Papua, we have shown that peripheral parasitemia is considerably higher in symptomatic patients with P. falciparum infections, compared with patients infected with P. vivax [20,30]. This may have contributed to the greater risk of severe thrombocytopenia in patients with P. falciparum infection (OR, 6.1), compared with those with P. vivax (OR, 3.7).
We have shown previously that the risk of anemia in this population is greatest in young patients, highlanders, and those presenting with recurrent episodes of malaria [3]. In contrast, in the current analysis, the risk of severe thrombocytopenia was significantly lower in patients with malaria who had had a prior episode of malaria within the preceding 2 months (OR, 0.8); this attenuation was most apparent in patients presenting with P. falciparum but not in those presenting with P. vivax monoinfection or without malaria. Furthermore, after the first year of life, lowland and non-Papuan patients had a low risk of severe thrombocytopenia. The risk of thrombocytopenia was significantly higher in Highland Papuans, and this was sustained throughout adulthood. Highlanders constitute an ethnic group originating from non-malaria-endemic regions who have not been under genetic selection pressure from malaria parasite infections. In the last decade, many highlanders have migrated at all ages to the lowland areas, where they have been exposed to malaria, often getting their first episodes of malaria in later life.
Our findings are consistent with lowland ethnicity or recent malaria resulting in a reduction of the host inflammatory response to acute malaria and decreased platelet activation and consumption. The pathogenic mechanisms by which platelets mediate disease severity remain to be delineated. However, clinical, autopsy, ex vivo, and in vitro studies have shown that platelets are involved in parasite sequestration [31], as well as in clumping and/or agglutination of infected and uninfected erythrocytes [32,33]. Platelets express Toll-like receptors (TLRs), which, on recognition of P. falciparum molecular patterns, release prepackaged inflammatory mediators [34]. This could partially explain the attenuation with repeat exposure, as repeated stimulation of TLRs leads to decreased signaling and decreased inflammatory responses [35]. Nitric oxide (NO) is also a key mediator of platelet homeostasis, and the decreased NO bioavailability found in both children [36] and adults [37] with severe and fatal malaria may contribute to increased platelet activation and consumption.
Our large-scale observational study has a number of limitations. First, platelet counts were only available in 26% of all presentations, and so there may be a degree of residual confounding in our multivariable analyses. However, the available hematology data rose to 80% in patients requiring admission. Although the risk of thrombocytopenia was greater in inpatients than outpatients, the magnitude of the other risk factors remained similar in both departments. Second, the surveillance program did not document the presence of all severe manifestations of malaria in these patients, so it is not possible in this data set to determine whether the presence of severe thrombocytopenia would have identified patients at risk of death in whom other World Health Organization (WHO) criteria for severe disease were not apparent [7]. Previous studies have used multivariate analysis to identify biomarkers predictive of poor outcome, but most have not included platelet counts [7]. Hence, it is possible that the mortality risk associated with severe thrombocytopenia may be better represented by other clinical biochemical and inflammatory markers. However, platelet counts are readily available from an automated blood count, a routine laboratory test that is widely accessible even in referral inpatient facilities and even some more remote health posts and that is more accessible than other recognized laboratory predictors of mortality in WHO severity criteria (such lactate and bicarbonate levels or creatinine level) [7]. Our study, the largest to date that examined relationships between severe thrombocytopenia and malaria mortality, highlights that severe thrombocytopenia should serve as a warning sign of poor outcome, particularly when coexisting with severe anemia. We believe that severe thrombocytopenia may be useful in guiding the need for referral or triage to a ward where a higher level of care is provided. Our analysis focused on applying a threshold of 50 000 platelets/μL, which was associated with an overall mortality of 3.9%, a PAF of 14.5% and sufficient power to determine other relevant confounding factors. However, the Abbreviations: CI, confidence interval; Hb, hemoglobin; OR, odds ratio. a Model presents risk with respect to the reference group of patients without severe anemia or thrombocytopenia (severe thrombocytopenia was defined as a platelet count of <50 000 platelets/μL, and anemia was defined as a Hb level of <5 g/dL). Adjusted for age group, sex, ethnicity, year, and recent malaria presentations in the past 2 months. b Model presents risk in patients with malaria, with respect to the reference group of patients without severe anemia or thrombocytopenia (severe thrombocytopenia was defined as a platelet count of <50 000 platelets/μL, and anemia was defined as a Hb level of <5 g/dL). Adjusted for species, sex, ethnicity, year, and recent malaria presentations in the past 2 months.
mortality risk rose to 7.9% in patients with a platelet count of <20 000 platelets/μL (5.6% in falciparum malaria and 3.6% in vivax malaria). We propose that a platelet count of ≤20 000 platelets/μL should be included as a defining severity criterion for both falciparum and vivax malaria. Prospective studies are warranted to evaluate the prognostic value of using platelet counts in conjunction with hemoglobin concentrations to define medical interventions and to determine the underlying processes by which thrombocytopenia contributes to the pathology of malaria.

Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.