Risk Factors for Severe Disease in Adults with Falciparum Malaria

Transcription

1 MAJOR ARTICLE Risk Factors for Severe Disease in Adults with Falciparum Malaria Anastasia Phillips, 1 Paul Bassett, 1 Sebastian Zeki, 1 Stanton Newman, 3 and Geoffrey Pasvol 1,2 1 Department of Infection and Tropical Medicine, Northwick Park Hospital, and 2 Wellcome Centre for Clinical Tropical Medicine, Wright Fleming Institute, St Mary s Campus, Imperial College London, and 3 Centre for Behavioural and Social Sciences in Medicine, University College London, London, United Kingdom Background. Over a 16-year period, we conducted a clinical study of malaria acquired worldwide in adults from malaria-nonendemic countries, to determine risk factors for severe Plasmodium falciparum malaria. Methods. All patients with confirmed malaria who were managed by our unit from 1991 to 2006 were prospectively evaluated. Factors predicting disease severity according to (1) strict World Health Organization (WHO) criteria, (2) a composite measure of unfavorable outcome, and (3) length of hospital stay were identified by logistic and linear regression analyses. Results. We evaluated 676 episodes of malaria, 482 (71%) due to P. falciparum and 194 (29%) due to nonfalciparum parasites. Black patients had a significantly reduced risk of developing WHO-defined severe falciparum malaria, with Asian patients having odds of severe falciparum malaria that were 8.05-fold (95% confidence interval [CI], fold) higher and white patients having odds that were 8.20-fold (95% CI, fold) higher. Black patients also had a reduced risk of an unfavorable outcome and of a prolonged stay in the hospital, compared with the risks for white or Asian patients. Of 6 patients with falciparum malaria who died, none were black. In univariate analysis, patients with parasitemias of 2% had odds of severe falciparum malaria 12-fold higher than those of patients with parasitemias of!2% (73% vs. 19%). Patients with a history of previous clinical malaria, regardless of ethnicity, had a significantly reduced risk of WHO-defined severe falciparum malaria (odds ratio, 0.35 [95% CI, ]). Conclusions. The findings of this study demonstrate that ethnicity and parasitemia are important independent risk factors for severe falciparum malaria in adults from malaria-nonendemic countries and that a history of previous clinical malaria significantly reduces the risk of WHO-defined severe falciparum malaria. Malaria is a major cause of morbidity and mortality worldwide, especially in young African children, and the risk of acquiring it is increasing among travelers, predominantly adults from countries in which malaria is not endemic. Children in Africa and travelers represent two distinct worlds of malaria [1]. Much research on malaria has appropriately been directed at children in areas of relatively high transmission; few studies have examined the full spectrum of malaria in travelers from malaria-nonendemic countries who had little or no exposure during childhood. Such studies Received 29 August 2008; accepted 3 December 2008; electronically published 25 February Reprints or correspondence: Prof. Geoffrey Pasvol, Dept. of Infection and Tropical Medicine, Imperial College London, Lister Unit, Northwick Park Hospital, Harrow, Middlesex HA1 3UJ, United Kingdom (g.pasvol@imperial.ac.uk). Clinical Infectious Diseases 2009; 48: by the Infectious Diseases Society of America. All rights reserved /2009/ $15.00 DOI: / can provide valuable information on a population group not complicated by other confounding factors, such as malnutrition, immunosuppression, and multiple pathologies. Those studies in travelers that have been published either have largely focused on demographic factors [2 4], subgroups of patients (e.g., those with malaria due to different parasite species) [5, 6], mild disease [7], severe disease alone [8], or children [9] or have been relatively small [10 12]. Studies in malaria-nonendemic regions provide a unique opportunity to examine the risk of severe disease in individuals of different ethnic backgrounds and relative to previous exposure. Few studies, however, have addressed this question comprehensively: one alluded to a lower frequency of severe disease in African patients than in other patients but did not include other defined ethnic groups or a control group with mild malaria [8]. A study in Italy suggested a lower frequency of severe disease in immigrants (1.3%) than in nonimmigrants Risk Factors for Severe Falciparum Malaria CID 2009:48 (1 April) 871

2 (9.2%) but did not provide details on ethnic background [2]. To examine the risk factors for severe malaria, we studied the clinical and laboratory features related to 676 episodes of malaria (482 due to Plasmodium falciparum) acquired worldwide and treated in a malaria-nonendemic setting in 3 distinct ethnic groups. METHODS Study participants. From April 1991 to May 2006, we identified all episodes of malaria diagnosed in adults (16 years or older) managed by the Infectious Diseases Unit at Northwick Park Hospital, London, United Kingdom. This specialist unit receives referrals from the emergency department, primary care physicians, and London s Heathrow Airport, as well as tertiary referrals. Data collection. Diagnoses were made on the basis of blood film or rapid diagnostic test results and were confirmed by the Health Protection Agency, Malaria Reference Laboratory, London, United Kingdom. Patients with malaria due to Plasmodium vivax, Plasmodium ovale, or Plasmodium malariae were grouped together as having nonfalciparum malaria. We used the World Health Organization (WHO) criteria for the definition of severe falciparum malaria (hereafter, WHOdefined severe falciparum malaria ) as the primary outcome variable [13]. However, to facilitate the assessment of risk factors (such as parasitemia and creatinine level) that form part of the WHO definition of severity as possible predictors, we developed 2 additional outcome variables. The composite variable unfavorable outcome was defined as death, admission to the intensive therapy unit (ITU), or a length of stay in the hospital of 5 days or more, and the continuous outcome variable length of hospital stay captured varying levels of disease severity. We used a length of stay of 5 days or more because 197% of our patients with nonfalciparum malaria were discharged before 5 days. Ethnicity was defined as black (black African and black Afro- Caribbean), Asian (from South Asia, specifically, India, Pakistan, Sri Lanka, and Afghanistan), and white. Ethnicity was self-declared by the patient or recorded by the admissions officer on the admission sheet. There were no patients of Southeast Asian or Melanesian origin. Residence was defined as the country in which the patient was resident during the preceding 12 months. Previous malaria was recorded only if patients clearly recalled previous clinical episodes; otherwise, unknown status was recorded. Latency was defined as the time, in days, between arrival in the United Kingdom and the onset of first symptoms. The study was approved by the Harrow Research Ethics Committee (REC 06/Q0405/24). Data analysis. In the descriptive analysis, categorical variables were summarized by frequencies, and numerical variables were summarized by means and SDs if normally distributed and by medians and interquartile ranges if nonnormally distributed. Logistic regression was used to examine the association between potential predictors and the outcome variables WHO-defined severe falciparum malaria and unfavorable outcome. Linear regression was used to examine the association between potential predictors and length of hospital stay. Only factors that showed evidence of a significant effect ( P!.02) in the univariate analyses were included in the multivariate analyses. A backward selection procedure was used to determine the initial model. Some numerical explanatory variables found to be highly skewed in their distribution were log transformed. Parasitemia was categorized according to a clinically relevant threshold (i.e.,!2% and 2%) parasitemias of!2% are treated with oral antimalarials and those of 2% with parenteral antimalarials, according to United Kingdom national guidelines [14]. Statistical analyses were performed using Stata software (version 9.2; StataCorp). RESULTS A total of 723 episodes of malaria were identified by blood film (720) or rapid diagnostic test (3); exclusions are shown in figure 1. Finally, 676 cases were analyzed, of which 482 (71%) were due to P. falciparum; the remainder (194 [29%]) were due to nonfalciparum parasites. P. falciparum was present in 4 of 6 mixed infections. Demographic and admission characteristics of patients with nonsevere and WHO-defined severe falciparum malaria are compared in table 1. No infections were acquired in Southeast Asia or Melanesia. In our study, 27 patients (24 African, 2 Asian, and 1 white) with falciparum malaria chose, for one reason or another, not to be admitted and were followed up by our infectious diseases assessment service. All the patients who were not admitted recovered, and none developed severe disease. Of the patients with severe malaria, 75 (16%) had WHO-defined severe falciparum malaria. Of these, 6 (8%) died, 6 (8%) had cerebral malaria, 31 (41%) had an impaired conscious level, 30 (40%) had parasitemias of 10%, 8 (11%) had renal failure, 7 (9%) developed acute respiratory distress syndrome (of whom 4 died), 1 (1%) had disseminated intravascular coagulation, and 1 (1%) had blackwater fever. A number of patients had combinations of these indicators of severity. Twenty nine (39%) of the patients with severe falciparum malaria were admitted to the ITU. Of the patients with severe cases, 55 (73%) were male, the majority (52/65 [80%]) resided in the United Kingdom, and most were holidaymakers in the malarial area. Only 23 of the patients with severe falciparum malaria were known to be taking antimalarials of any kind; of these antimalarials, only 8 were taken as prescribed, and 6 of these 8 were inappropriate. Thus, only 2 (3%) of 75 patients with severe disease had received a regimen regarded as adequate. Of the 6 patients who died, 4 were Asian, 2 were white, and none was black. 872 CID 2009:48 (1 April) Phillips et al.

3 Figure 1. Patient selection and exclusions. Unfavorable outcome was defined as death, admission to the intensive therapy unit, or length of stay in the hospital of 5 days or more. WHO, World Health Organization. None received adequate antimalarials 3 took no antimalarials, and 3 had inadequate regimens. Those with severe falciparum malaria were found to be older ( P p.002), with an increase in age of 10 years being associated with the odds of severe malaria increasing by 130%. White patients had the highest risk of severe falciparum malaria, and black patients had the lowest ( P!.001). There was no significant difference in severity between Asian and white patients ( P p.21). The lowest frequency of severe falciparum malaria occurred among those who had traveled to West Africa, and the highest frequency occurred among those who had traveled to southern Africa ( P!.001). Any history of previous malaria was associated with a significant decrease in the likelihood of severe falciparum malaria ( P!.001), with the odds of severe malaria being 5 times lower for those with any history of previous malaria. Overall, 50 (71%) of the patients with severe disease had not had malaria before, 5 had had it once only, and 10 had had it 13 times; for the remaining 5, their status was unknown. Even patients who reported between 1 and 3 attacks showed a reduced likelihood (5%) of developing WHOdefined severe falciparum malaria, compared with those who had not ever had clinical malaria (31%). Both parasitemia ( 10%) and creatinine level ( 265 mmol/ L) are part of the definition of severe malaria, and the significance of these associations with severity needs to be interpreted with caution. It is, however, important that 27% of patients classified as having severe falciparum malaria by the WHO criteria had parasitemias of!2% but that only 19% of those with parasitemias of 2% did not have severe falciparum malaria ( 2% was chosen as the cutoff because it is a determining factor in the mode of treatment given in the United Kingdom guidelines [14]). A number of other clinical and laboratory measures including pulse and respiratory rate; hemoglobin level; WBC, neutrophil, and platelet counts; calcium level; creatinine level; albumin level; bilirubin level; alanine aminotransferase level; activated partial thromboplastin time; and C-reactive protein level were associated with an increased risk of severe falciparum malaria (table 1). Importantly, ethnic group, lack of a history of previous malaria, decreased hemoglobin level and platelet count, and increased WBC count remained significantly associated with WHO-defined severe falciparum malaria in a multivariate logistic regression analysis (table 2). The odds of severe falciparum malaria were as much as 8-fold higher among white Risk Factors for Severe Falciparum Malaria CID 2009:48 (1 April) 873

4 Table 1. Characteristics of case patients with falciparum malaria, with severity defined according to World Health Organization criteria. Falciparum malaria Characteristic All (n p 482) Nonsevere (n p 407) Severe (n p 75) OR (95% CI) P Age, a mean SD, years ( ).002 Sex.25 Female 156 (33) 136 (33) 20 (27) 1.00 (reference) Male 326 (67) 271 (67) 55 (73) 1.38 ( ) Ethnic group!.001 Black 309 (64) 293 (72) 16 (21) 1.00 (reference) Asian 94 (20) 67 (16) 27 (36) 7.38 ( ) White 79 (16) 47 (11) 32 (43) 12.5 ( ) Area visited!.001 West Africa 288 (60) 256 (63) 28 (37) 1.00 (reference) East/Central Africa 135 (28) 108 (27) 26 (35) 2.20 ( ) Southern Africa 25 (5) 19 (5) 16 (15) 5.29 ( ) South Asia 33 (7) 24 (6) 10 (13) 3.81 ( ) Residence.50 United Kingdom 285 (77) 233/306 (76) 52/65 (80) 1.00 (reference) Other 86 (23) 73/306 (24) 13/65 (20) 0.80 ( ) No. of days ill, median (IQR) 4 (2 7) 4 (2 6) 5 (3 7) 1.02 ( ).47 Latency, b median (IQR), days 9 (4 14) 10 (4 15) 8 (1 12) 0.68 ( ).002 Previous malaria No 160 (39) 110/345 (32) 50/70 (71) 1.00 (reference) Yes 255 (61) 235/345 (68) 20/70 (29) 0.19 ( ) Receipt of antimalarials.35 No 294 (64) 244/389 (63) 50/73 (68) 1.00 (reference) Yes 168 (36) 145/389 (37) 23/73 (32) 0.44 ( ) Parasitemia!2% 331 (81) 20 (27) 1.00 (reference) 2% 76 (19) 55 (73) 12.0 ( ) Temperature, mean SD, C ( ).66 Pulse rate, a mean SD, beats/min ( )!.001 Blood pressure, mean SD, mm Hg Systolic a ( ).32 Diastolic a ( ).14 Respiratory rate, a mean SD, breaths/min ( )!.001 Hemoglobin level, mean SD, g/dl ( )!.001 Cell counts, median (IQR), 10 9 cells/l WBCs 5.4 ( ) 5.3 ( ) 6.5 ( ) 1.33 ( )!.001 Neutrophils 3.7 ( ) 3.5 ( ) 4.9 ( ) 1.45 ( )!.001 Lymphoctyes 1.0 ( ) 1.0 ( ) 0.9 ( ) 0.91 ( ).60 Monocytes 0.4 ( ) 0.5 ( ) 0.4 ( ) 0.42 ( ).06 Platelets c 102 (64 155) 115 (76 164) 44 (29 74) 0.25 ( )!.001 Glucose level, a median (IQR), mmol/l 6.0 ( ) 6.0 ( ) 5.9 ( ) 1.62 ( ).35 Corrected calcium level, d mean SD, mmol/l ( )!.001 Sodium level, a mean SD, mmol/l ( )!.001 Creatinine level, e median (IQR), mmol/l 99 (84 114) 98 (84 111) 108 (82 164) 10.0 ( )!.001 Albumin level, a mean SD, g/l ( )!.001 Bilirubin level, e median (IQR), mmol/l 23 (16 37) 21 (15 33) 46 (25 79) 6.15 ( )!.001 Alanine aminotransferase level, e median (IQR), IU/L 40 (26 63) 38 (25 58) 50 (33 101) 2.41 ( )!.001 Activated partial thromboplastin time, a mean SD, s ( ).02 C-reactive protein level, c median (IQR), mg/l 100 (55 154) 96 (50 144) 130 (82 230) 1.49 ( )!.001 NOTE. Data are number or proportion (%) of patients, unless otherwise indicated. Boldface type indicates statistical significance. To convert values for glucose level to mg/dl, divide by ; to convert values for calcium level to mg/dl, divide by 0.250; to convert values for creatinine level to mg/dl, divide by 88.4; to convert values for albumin level to g/dl, divide by 10; and to convert values for bilirubin level to mg/dl, divide by IQR, interquartile range. a ORs are given for a 10-U change in the explanatory variable. b Time between arrival in the United Kingdom and onset of the first symptoms. c ORs are given for a 50-U change in the explanatory variable. d OR is given for a 0.1-U change the explanatory variable. e Variable was analyzed on a log scale.!.001!.001

5 Table 2. Multivariate analysis of factors associated with the risk of severe falciparum malaria, according to the World Health Organization definition of severity. Variable OR (95% CI) P Ethnic group!.001 Black 1.00 (reference) Asian 8.05 ( ) White 8.20 ( ) Previous malaria.01 No 1.00 (reference) Yes 0.35 ( ) Parasitemia!.001!2% 1.00 (reference) 2% 4.93 ( ) Hemoglobin level 0.75 ( ).002 WBC count 1.39 ( ).001 Platelet count a 0.47 ( )!.001 Creatinine level b 13.1 ( ).001 NOTE. The final regression model was based on 408 patients with available data (of a total of 482 patients). a OR is given for a 50-U change in the explanatory variable. b Variable was analyzed on a log scale. and Asian patients than among black patients, even after adjustment for other risk factors. Moreover, in this multivariate analysis the odds of severe disease was significantly reduced ( 3-fold) among patients with any history of previous malaria relative to those among patients who had not ever had malaria. Although parasitemia is included in the WHO definition (with a cutoff of 10%), it is of note that, when a cutoff of 2% was used, parasitemia had a direct relationship with severe disease, with the odds of severe falciparum malaria in multivariate analysis being almost 5 times higher among patients with parasitemias of 2%. A reduced hemoglobin level and platelet count and a raised WBC count were associated with severity. A 1-U increase in WBC count ( cells/l, or 1000 cells/ml) increased the odds of severe falciparum malaria by 39%. To exclude the confounding that results from parasitemia and creatinine level being included in the WHO definition of severity, we analyzed the cases according to alternate definitions of severity. An unfavorable outcome (as defined in Methods) occurred in 127 patients (26%) (figure 1). Univariate analysis revealed associations similar to those found in the analysis using the WHO definition of severity (table 1). Multivariate analysis once again indicated that ethnic group had a significant independent association with the occurrence of an unfavorable outcome, along with other factors shown in table 3. The odds of an unfavorable outcome were 4 times higher among white and Asian patients than among black patients. Patients with parasitemias of 2% had an increased likelihood of an unfavorable outcome, with odds 14 times higher than those of patients with parasitemias of!2%. Other significant clinical and laboratory factors are shown in table 3. It is of note that previous exposure was not significant in the multivariate analysis of an unfavorable outcome. Multiple linear regression was used to analyze the effect that the variables have on a continuous outcome measure, namely, length of hospital stay (table 4). After adjustment for the other variables in the model, there was a significant association between ethnic group, parasitemia, temperature, respiratory rate, neutrophil count, platelet count, albumin level, and length of hospital stay. As in the univariate analysis, black patients had the shortest hospital stay relative to the other 2 ethnic groups. The lengths of stay in the hospital for white and Asian patients were, on average, 40% and 34% longer than that for black patients, respectively. Parasitemia was again found to have a large effect on the length of hospital stay. Patients with parasitemias of 2% 9.9% had stays in the hospital that were, on average, 32% longer than those of patients with parasitemias of!2% and 67% longer than those of patients with parasitemias of 10%. Increased values for temperature, respiratory rate, and neutrophil count on admission were also associated with longer stays in the hospital. A 10-U increase in respiratory rate was associated with the length of hospital stay increasing by 18%. Last, reduced platelet and albumin values were associated with a longer stay in the hospital. Finally, for all 3 measures of disease severity, no significant interactions between ethnic group and any of the other variables were found. In particular, there was no interaction with history of previous malaria, implying that the ethnic differences were independent of previous exposure. Table 3. Multivariate analysis of factors associated with an unfavorable outcome of falciparum malaria. Variable OR (95% CI) P Ethnic group!.001 Black 1.00 (reference) Asian 4.78 ( ) White 3.88 ( ) Parasitemia!.001!2% 1.00 (reference) 2% 4.38 ( ) WBC count 1.17 ( ).04 Monocyte count 0.25 ( ).01 Albumin level a 0.47 ( ).003 Bilirubin level b 1.91 ( ).005 NOTE. Unfavorable outcome was defined as death, admission to the intensive therapy unit, or length of stay in the hospital of 5 days or more. The final regression model was based on 436 patients with available data (of a total of 482 patients). a OR is given for a 10-U change in the explanatory variable. b Variable was analyzed on a log scale. Risk Factors for Severe Falciparum Malaria CID 2009:48 (1 April) 875

6 Table 4. Multivariate analysis of factors associated with length of hospital stay in patients with falciparum malaria. Variable OR (95% CI) P Ethnic group!.001 Black 1.00 (reference) Asian 1.40 ( ) White 1.34 ( ) Parasitemia!.001!2% 1.00 (reference) 2% 9.9% 1.32 ( ) 10% 1.67 ( ) Temperature 1.08 ( ).004 Respiratory rate a 1.18 ( ).008 Neutrophil count b 1.03 ( ).04 Platelet count a 0.90 ( )!.001 Albumin level b 0.86 ( ).01 NOTE. The final regression model was based on 375 patients with available data (of a total of 476 patients; 6 patients who died were excluded). a ORs are given for a 10-U change in the explanatory variable. b ORs are given for a 1-U change in the explanatory variable. DISCUSSION The present large study, which included 482 patients with falciparum malaria, examined risk factors for 3 different measures of disease severity. Patients belonged to 3 distinct ethnic groups, and although a number of black and Asian patients had had malaria previously, the majority were second-generation United Kingdom residents without previous exposure. We were thus able to distinguish between the effects of innate and acquired immunity in relation to the development of severe disease. The most striking observation was that black patients had a significantly reduced risk of developing severe disease by all 3 of the measures. When the WHO definition was used, the odds of severe falciparum malaria in multivariate analysis were 18 times higher among white and Asian patients relative to those among black patients. Numerous single parameters correlated with WHO-defined severe falciparum malaria in univariate analysis. In multivariate analysis, white or Asian ethnicity, together with parasitemia of 2% and no history of previous malaria, was the most predictive of WHO-defined severe falciparum malaria, and similar variables (with the exception of previous exposure) were predictive of an unfavorable outcome. In the multivariate analysis of length of stay, there was also a highly significant effect of ethnic group and parasitemia. Black patients had the shortest stay in the hospital; durations were 34% longer among Asian patients and 46% longer among white patients than among black patients. Patients with parasitemias of 2% stayed in the hospital at least 50% longer than did those with parasitemias of!2%. Having established that black patients were at reduced risk of severe falciparum malaria, a major consideration was whether this was due to either innate or acquired immunity. Ethnicity was found to be independent of the other risk factors, and among each ethnic group the proportion of patients with WHO-defined severe falciparum malaria was greater for those who did not have a history of previous malaria than for those who did. Even patients who reported 1 3 episodes had a reduced likelihood of WHO-defined severe falciparum malaria, compared with those who had not ever had clinical malaria. Our data provide the first demonstration in adults outside a malaria-endemic country that immunity to disease may be acquired after relatively little exposure. Similar findings have suggested the relatively rapid acquisition of protection against severe malaria after 1 or 2 infections, albeit in children living in an malaria-endemic area [15]. Although we have shown that a history of malaria might protect against severe disease, presumably via mechanisms of acquired immunity, other factors most likely genetic that contribute to protection remain. However, even in Africa the sickle hemoglobin gene, as important as it may seem, probably accounts for only 2% of the total variation in the overall effect that host genetics has on the risk of malaria [16]. It is important to point out that previous exposure was not found to be significantly associated with an unfavorable outcome or length of hospital stay. Contrary to our findings, a recent study reported that patients (largely black Africans) originating from malaria-endemic countries were not less likely to develop severe disease [17]. However, the definition of severe malaria is crucial; a lessstringent modification of the WHO definition was used in that study, resulting in a relatively small number (25) of severe cases. By the WHO criteria, only 10 of the cases could be described as being truly severe. Of these, 5 (50%) occurred in white or Asian patients, who were thus overrepresented, given that they constituted only 33% of the overall study population. In a recent study of 33 deaths due to falciparum malaria in Switzerland, the case fatality ratio was 18 times higher among Europeans than among non-europeans [18]. Our study contained 3 distinct ethnic groups, 2 of which (Asians and white patients) were found to be at increased risk of severe malaria. Regarding the limitations of our study, we are aware that the WHO criteria for severe disease include a number of unvalidated parameters [19]. Although unrousable coma clearly identifies a group at risk, it has been debated whether an alteration in consciousness of lesser severity is a truer measure of severe disease. We therefore conducted analyses including and excluding those patients with impaired consciousness alone as a criterion for severe disease, and both found the same relative risk for the different ethnic groups (data not shown). Likewise, there is little agreement as to what level of parasitemia indicates severe malaria (e.g., 5% or 10%) we chose the morestringent definition of 10%. Moreover, although age was found to be a significant risk factor for WHO-defined severe 876 CID 2009:48 (1 April) Phillips et al.

7 falciparum disease in univariate analysis (in agreement with the findings of a recent study [20]), it was not a significant risk factor in multivariate analysis. In addition, our study, conducted as part of routine clinical care, was not able to capture in any significant level of detail the amount of time that had elapsed since a previous clinical episode, which we believe may be an additional important factor in determining the risk of severe disease. A further consideration for the use of the WHO definition of severe malaria in descriptive clinical studies is that it was developed largely for the purpose of standardizing clinical trials and reflects admission criteria rather than outcome measures. In clinical practice, many of the measures of severity are not present on admission and may develop insidiously or sometimes unexpectedly during the subsequent course of disease [8]. These delayed manifestations are commonly those of acute respiratory distress syndrome, acute renal failure, and death. There is the additional problem that parasitemia and creatinine level are included in the WHO definition, thus excluding their usefulness as determinants of severe malaria. Therefore, to increase the robustness of our conclusions we also analyzed our patients by means of an additional measure, unfavorable outcome (death, admission to the ITU, or length of stay in the hospital of 5 days or more). Additionally, because dichotomizing variables such as length of hospital stay has its own inherent problems [21], we also used length of stay as a simple continuous measure of disease severity. Our study provides further support for the cutoff of 2% for parasitemia in the United Kingdom clinical guidelines for the use of parenteral antimalarials [22]. In our study, patients with parasitemias of 2% had odds of severe malaria 12-fold higher than those of patients with parasitemias of!2%. A recent systematic review of the treatment of malaria in the United States advocates parenteral treatment only when parasitemias exceed 5% [23]. Given that a randomized, controlled trial of outcome among nonimmune patients is unlikely to be conducted in the future, we would recommend the use of the more-conservative threshold of 2%. Intravenous artesunate would now be preferable [24], although all patients enrolled in the present study were treated with quinine. In conclusion, our study conducted in a malaria-nonendemic country and of sufficient size, ethnic mix, and range of disease severity demonstrates that ethnicity and parasitemia are major independent risk factors for severe falciparum malaria. In the clinical assessment of any case of malaria, a number of findings relating to prognosis need to be taken into account, and we have also shown that the risk associated with ethnicity is independent of other factors and might be the result of known (or unknown) genetic polymorphisms. Our data, generated using the WHO definition of severe falciparum malaria, support the contention that a history of previous clinical malaria significantly reduces the likelihood of severe disease, but determining how this finding and the finding of a variation in susceptibility among ethnic groups translate into management guidelines will require further targeted prospective studies. Acknowledgments We extend our sincere thanks to all the patients who participated in this study. We also thank the scientists Dr. Barbara Clough, Julie Black, and Abiola Atilola as well as Nicolette Davies and the numerous junior doctors who conscientiously contributed to the maintenance of an up-to-date database over the years. Financial support. Wellcome Trust (numerous grants to G.P.); Northwest London Hospitals NHS Trust (to G.P.). Potential conflicts of interest. All authors: no conflicts. References 1. Wellems TE, Miller LH. Two worlds of malaria. N Engl J Med 2003; 349: Matteelli A, Colombini P, Gulletta M, Castelli F, Carosi G. Epidemiological features and case management practices of imported malaria in northern Italy Trop Med Int Health 1999; 4: Jelinek T, Schulte C, Behrens R, et al. Imported falciparum malaria in Europe: sentinel surveillance data from the European Network on Surveillance of Imported Infectious Diseases. Clin Infect Dis 2002;34: Strauss R, Pfeifer C. Malaria in Austria Euro Surveill 2003; 8: Kochar DK, Saxena V, Singh N, Kochar SK, Kumar SV, Das A. Plasmodium vivax malaria. Emerg Infect Dis 2005; 11: Bottieau E, Clerinx J, Van Den Enden E, et al. Imported non-plasmodium falciparum malaria: a five-year prospective study in a European referral center. Am J Trop Med Hyg 2006; 75: Svenson JE, MacLean JD, Gyorkos TW, Keystone J. Imported malaria: clinical presentation and examination of symptomatic travelers. Arch Intern Med 1995; 155: Bruneel F, Hocqueloux L, Alberti C, et al. The clinical spectrum of severe imported falciparum malaria in the intensive care unit: report of 188 cases in adults. Am J Respir Crit Care Med 2003; 167: Ladhani S, Aibara RJ, Blaze M, Smith V, Shingadia D. Trends in imported childhood malaria in the UK: Arch Dis Child 2006; 91: Kain KC, Harrington MA, Tennyson S, Keystone JS. Imported malaria: prospective analysis of problems in diagnosis and management. Clin Infect Dis 1998; 27: Singh K, Wester WC, Trenholme GM. Problems in the therapy for imported malaria in the United States. Arch Intern Med 2003;163: Miura T, Kimura M, Koibuchi T, et al. Clinical characteristics of imported malaria in Japan: analysis at a referral hospital. Am J Trop Med Hyg 2005; 73: World Health Organization. Severe falciparum malaria: World Health Organization, Communicable Diseases Cluster. Trans R Soc Trop Med Hyg 2000; 94(Suppl 1):S Lalloo DG, Shingadia D, Pasvol G, et al. UK malaria treatment guidelines. J Infect 2007; 54: Gupta S, Snow RW, Donnelly CA, Marsh K, Newbold C. Immunity to non-cerebral severe malaria is acquired after one or two infections. Nat Med 1999; 5: Mackinnon MJ, Mwangi TW, Snow RW, Marsh K, Williams TN. Heritability of malaria in Africa. PLoS Med 2005; 2:e Jennings RM, De Souza JB, Todd JE, et al. Imported Plasmodium falciparum malaria: are patients originating from disease-endemic areas less likely to develop severe disease? A prospective, observational study. Am J Trop Med Hyg 2006; 75: Risk Factors for Severe Falciparum Malaria CID 2009:48 (1 April) 877

8 18. Christen D, Steffen R, Schlagenhauf P. Deaths caused by malaria in Switzerland Am J Trop Med Hyg 2006; 75: Anstey NM, Price RN. Improving case definitions for severe malaria. PLoS Med 2007; 4:e Dondorp AM, Lee SJ, Faiz MA, et al. The relationship between age and the manifestations of and mortality associated with severe malaria. Clin Infect Dis 2008; 47: Altman DG, Royston P. The cost of dichotomising continuous variables. BMJ 2006; 332: Lalloo DG, Trevett AJ, Paul M, et al. Severe and complicated falciparum malaria in Melanesian adults in Papua New Guinea. Am J Trop Med Hyg 1996; 55: Griffith KS, Lewis LS, Mali S, Parise ME. Treatment of malaria in the United States: a systematic review. JAMA 2007; 297: Dondorp A, Nosten F, Stepniewska K, Day N, White N. Artesunate versus quinine for treatment of severe falciparum malaria: a randomised trial. Lancet 2005; 366: CID 2009:48 (1 April) Phillips et al.