Obesity and Diabetes as Risk Factors for Severe Plasmodium falciparum Malaria: Results From a Swedish Nationwide Study

Summary In this nationwide observational study of 937 adults diagnosed with Plasmodium falciparum malaria in Sweden, Charlson comorbidity score ≥1 as well as diabetes and obesity were significantly associated with severe malaria in both nonimmune travelers and immigrants from endemic countries.

Obesity and noncommunicable diseases (NCDs), such as diabetes, hypertension, and cardiovascular disease, have increased globally, including in malaria-endemic regions [1]. In addition, a significant proportion of travelers are older [2], and an estimated one-third of travelers to malarious countries have underlying medical conditions [3]. This changing disease panorama in populations at risk of malaria warrants the need to establish how comorbidities affect severity of malaria.
Studies on comorbidities have largely focused on coinfections, with human immunodeficiency virus (HIV) and chronic hepatitis B reported to influence the risk of severe malaria [4,5]. The role of NCDs has only been assessed in a few studies. Overweight was observed to affect disease course in uncomplicated cases in Thailand [6], and higher prevalence of asymptomatic Plasmodium falciparum infections was found among individuals with type 2 diabetes in Ghana [7].
Old age has been identified as a risk factor for both severe and fatal malaria in travelers [8][9][10][11], and longer hospital stay was reported in patients aged ≥65 years with chronic diseases [12]. However, no study has systematically assessed how comorbidities affect severity of malaria. Such evaluation is important to improve public health strategies and support clinicians to recognize patients at risk.
In this nationwide observational study of imported malaria in Sweden over 20 years, we assessed if comorbidity, and any chronic condition in particular, is associated with severe malaria in adults diagnosed with P. falciparum.
in Sweden with mandatory reporting by diagnosing clinicians and microbiology laboratories, providing high detection sensitivity [13]. All adults (≥18 years of age) with a first episode of microbiologically confirmed P. falciparum and complete medical records from 1 January 1995 to 12 April 2013 were included, and for Umeå until 31 August 2013 and Stockholm until 31 May 2015. Patients without symptoms and asexual parasites after treatment elsewhere were excluded.

Data Collection
Medical records from identified cases were provided by 18 hospitals managing malaria in Sweden. Data were retrieved regarding sociodemographics, travel history, chemoprophylaxis, clinical presentation, comorbidities, patient and healthcare delay (days from symptoms onset until healthcare contact, and from healthcare contact to diagnosis), intensive care, duration of hospital stay, treatment, and outcome, as well as routine blood chemistry and microbiology data including parasitemia, HIV, and hepatitis status. Data on weight and height were collected in Stockholm and Umeå. Medication lists and previous International Classification of Diseases (ICD) codes in electronic medical records were reviewed to capture additional chronic diseases.

Malaria Diagnosis
Malaria was diagnosed by microscopy of thick and thin blood films stained with Giemsa or Field stain. Parasite species were occasionally determined by polymerase chain reaction. Parasitemia (percentage of infected erythrocytes) was estimated in thin smears, or by counting parasites in thick films either against leukocytes or ocular fields.

Primary Outcome
Severe malaria was defined using 2012 World Health Organization (WHO) criteria [14], with minor modifications [15] (Table 1), and hyperparasitemia >5% according to previous WHO definition [16]. A sensitivity analysis without hyperparasitemia as single criterion was also performed.

Covariables
Medical conditions were categorized according to the ICD, Tenth Revision (ICD-10). Comorbidity was assessed both as individual diagnoses and as severity-weighted scores using the Charlson comorbidity index adjusted to ICD-10 [17], and with HIV without AIDS given a score of 1 [18]. Only chronic diseases present at the time of malaria diagnosis and history of malignancies were included in the analysis. Previous resolved conditions such as pneumonia or appendectomy were not included.
For patients with weight and height recorded at time of malaria diagnosis, body mass index (BMI) was calculated as weight in kilograms divided by square height in meters, and categorized according to WHO's BMI classification for adults [19]. Obesity was defined as BMI ≥30 (WHO obesity class I-III).
Metabolic syndrome was defined according to the International Diabetes Federation as BMI ≥30 together with 2 additional metabolic risk factors: diabetes, dyslipidemia, and/or hypertension [20].
Individuals born in countries with high malaria transmission in sub-Saharan Africa [21] were referred to as of "endemic origin, " and all others as "non/low endemic origin. " Duration of residency in a malaria-free country for patients of endemic origin was categorized as <15 and ≥15 years, based on previous findings [22].

Statistical Analysis
Statistical analyses were performed using Stata version 13 software (StataCorp). Categorical data were compared using χ 2 or Fisher exact test, and continuous data using Wilcoxon-Mann-Whitney test. Univariable and multivariable logistic regression were used to assess if comorbidity was associated with severe malaria. Age and endemic origin were included as possible confounders in all multivariable analyses based on biological plausibility. Additional patient characteristics affecting severity in univariable analysis (with P < .20) were included in the multivariable model to further adjust for confounding. Factors not associated with severity (P > .05) and not changing the effect measure of the main exposures were subsequently excluded. Age was included as continuous variable, after confirming linearity. Potential interactions were tested between variables in the final model. Maximum likelihood ratio test was used to determine best model fit. To account for missing BMI, a multiple imputation model with chained equations was performed based on variables related to severe malaria, obesity status, and missing BMI.

Patient Characteristics
In total, 937 adults with P. falciparum malaria were included, representing 71.9% (937/1304) of all notified P. falciparum cases in adults during the study period ( Figure 1). Median age was 37 years (range, 18-83 years), and most were male (66.5%). Five hundred forty-seven patients (58.4%) originated in sub-Saharan Africa; 441 (80.6%) were Swedish residents and 98 (17.9%) newly arrived immigrants or temporary visitors. Among the 388 patients of non/low endemic origin, 342 (88.1%) were from Sweden. Infections were predominantly acquired in Western and Eastern Africa ( Table 2). Ninety-two patients (9.8%) fulfilled the severe malaria definition, and 68 (7.3%) had severe criteria without hyperparasitemia as single criterion. One fatal outcome was reported in a Swedish man aged 43, corresponding to a case fatality ratio of 0.1%.
Healthcare delay, age, and non/low endemic origin were associated with severe malaria and included in the multivariable model.  (Table 3), with no effect modification by age (P = .46).
Individual diagnoses associated with severe malaria in the univariable analysis were diabetes, hypertension, cardiovascular disease, and HIV (Table 3). After adjustment for age, healthcare delay, and patient origin, the associations remained significant for diabetes (aOR, 2.98 [95% CI, 1.25-7.09]) and HIV (aOR, 5.37 [95% CI, 1.71-16.86]). There was no overlap between HIV and diabetes; moreover, HIV resulted in a general drift in ORs, likely explained by small sample bias or noncollapsibility, thus not included in the final model.
Additional adjustment for hypertension and cardiovascular disease did not affect the association between diabetes and severity (OR, 2.75 [95% CI, 1.13-6.70]). Four of 33 diabetic patients had type 1 diabetes and the only type 1 diabetic patient with severe malaria had BMI 32.7, hypertension, and hyperlipidemia.
The only detected interaction was between healthcare delay and diabetes, but as the interaction was weak (P = .04) and based on only 6 patients, it was not included in the multivariable model.

BMI and Severe Malaria
Data on BMI were available for 219 of the 569 patients from Stockholm and Umeå. Patients with obesity were older, more often of endemic origin, and had more comorbidities (especially diabetes and hypertension), but were similar to nonobese patients regarding sex, chemoprophylaxis, patient and healthcare delay, and none with HIV (Supplementary Table 2 Table 4). Only 5 of 219 patients fulfilled the complete metabolic syndrome criteria (3 with severe malaria), thus were too few to analyze. In the subset with collected BMI data, diabetes was associated with severe malaria (aOR, 3.37 [95% CI, 1.19-9.54]), and adjusting for obesity reduced the aOR to 2.39 (95% CI, .61-9.32).

Patient Origin and Time in Nonendemic Country
Obesity and diabetes were most prevalent among patients of endemic origin with long residency in Sweden: 16.5% (15/91),    [14] without hyperparasitemia as single criterion. One individual without information on severe signs but known hyperparasitemia >5% excluded in analysis. among patients of endemic origin with residency ≥15 years (obesity: aOR, 6.88 [95% CI, 1.21-39.24]; diabetes: aOR, 5.32 [95% CI, 1.01-28.18]). Interaction analysis could not verify that patient origin, and time of residency modified the association of either diabetes or obesity with severity (all P > .50).

DISCUSSION
In this nationwide study including 937 adults with P. falciparum malaria in Sweden, we identified comorbidity, and specifically diabetes, obesity, and components of the metabolic syndrome, as risk factors for severe malaria both in nonimmune travelers and immigrants from sub-Saharan Africa.
Patients with ≥2 chronic diseases or a Charlson score ≥1 had an increased risk of severe malaria. Moreover, age, healthcare delay, nonendemic origin, and HIV were associated with severe malaria as previously shown [4,8,11]. However, few patients were HIV infected, and the association with severity became nonsignificant when excluding hyperparasitemia as single criterion for severe malaria.
Diabetes, hypertension, and cardiovascular disease were all associated with severe malaria in the univariable analysis; however, diabetes was the only NCD that on its own remained significant after adjustments. Only a few had type 1 diabetes, so conclusions can only be drawn for type 2 diabetes. Other NCDs might also affect the outcome of malaria; however, prevalence in these travelers was too low to assess their impact.
Obesity was highly associated with severity in the population subset with retrieved BMI data. Type 2 diabetes is a complication of obesity and the 2 metabolic disorders often coexist. Adjusting for obesity somewhat reduced the odds of severe malaria among diabetics, whereas adjusting for the potential intermediate factors diabetes, hypertension, and cardiovascular disease [23] did not substantially change the association between obesity and severity. Having at least 1 metabolic risk factor was associated with increased odds for severe malaria, and a combination including obesity conferred even higher odds.
Reports on NCDs and malaria are few. To our knowledge, no study has assessed the association between diabetes and severe malaria. Two reports from Ghana have shown that semi-immune adults with type 2 diabetes were more susceptible to Plasmodium infection than controls [7,24], but with no inference on severity. A recent animal study indicates enhanced transmission of parasites from diabetics to mosquitoes [25]. BMI in relation to malaria has mainly been assessed in populations with considerably lower BMI [6,26,27]. Underweight was identified as a protective factor against severe malaria [26],whereas overweight was associated with progression to severe malaria after treatment start [6], and fatal cases had higher BMI (mean, 25.3) compared with nonfatal severe cases (20.4) [28]. However, none of these studies investigated the effect of obesity. Here, median BMI was higher among severe cases (29.3) than nonsevere cases (24.7); and obesity (BMI ≥30), but not overweight (BMI [25][26][27][28][29], was strongly associated with severe malaria at diagnosis. Obesity and diabetes (type 1 and 2) are well recognized to increase severity of infections [29,30]. Both conditions are characterized by low-grade chronic inflammation and altered levels of nutrients and metabolic hormones with immunomodulatory effects [30,31]. Equally important, metabolic changes could have specific effects on the malaria parasite and pathogenesis. Acute malaria is well recognized to influence plasma glucose and lipid levels [14,32,33]. Parasite growth in vitro is affected by glucose levels [34]; in Ghana, diabetic adults had 5% greater risk of asymptomatic parasitemia for each millimolar increase of plasma glucose [7]. Moreover, lipoproteins are important for parasite cell membranes and endothelial adherence of infected erythrocytes [35,36].
Obese patients had higher parasitemia compared with nonobese patients (especially >10%), a feature also described in obese mice [37], and hyperparasitemia >5% tended to be more common in diabetics. Nonetheless, approximately half of diabetic patients and obese patients with severe malaria had severe criteria without hyperparasitemia, and the associations with severity were even stronger when hyperparasitemia was excluded as single severe criterion, suggesting that parasitemia could not solely explain the more severe presentations. The mechanisms by which these comorbidities affect malaria pathology clearly need to be further investigated.
Numerous studies have shown that immigrants from high-endemic regions are at lower risk of severe and fatal malaria compared to nonimmune travelers [9][10][11]. Recently, we found that immigrants from sub-Saharan Africa residing ≥15 years in Sweden had a similar risk of severe malaria as nonimmune travelers [22]. Interestingly, both obesity and diabetes were most prevalent in this group. Poor antibody response after vaccination toward bacterial and viral infections has been observed in obese individuals [31]; hence, lifestyle diseases such as diabetes and obesity might possibly affect maintenance of immunity acquired against severe malaria. Although the association with severe malaria was strongest among patients of endemic origin with residency ≥15 years, larger studies are needed to truly assess possible interaction. Our study has several limitations. The retrospective design could lead to misclassification of exposures and outcome variables. The current data were entered as part of a larger epidemiological study without priori hypotheses. Moreover, in Sweden, fever after tropical visits is managed by infectious diseases specialists using standardized protocols. A high sensitivity of comorbidity assessment was achieved by using medical records, registered ICD codes, and medication lists. Moreover, prevalence data on chronic diseases and BMI present before the malaria episode produce an acceptable proxy for longitudinal data. However, a majority of the HIV tests were taken in conjunction with the malaria episode and severe cases were more often tested, which could imply a detection bias for HIV. Nonetheless, 18 of 21 HIV cases were known before, and among the 3 newly detected, none had severe malaria. Data on BMI were lacking from a large proportion of patients. Incorporating the factors associated with missing data (Supplementary Table 2) in a multiple imputation model resulted in similar effect of obesity on severity as the complete case analysis. Systematic health screening including BMI and glucose control, preferably in prospective studies, are needed to confirm our results. Possibly, some patients had yet undiagnosed chronic diseases, such that nondifferential misclassification might have diluted the effects.
The strength of our study is that it includes clinical data from a large population of travelers, with a mixed constitution of individuals previously exposed and unexposed to malaria, diagnosed at multiple centers. Case detection was based on the national surveillance system with high detection sensitivity [13]. The results are well generalizable to settings with a similar spectrum of imported malaria.
We believe our findings have implications beyond the management of malaria in travelers. Sub-Saharan Africa is facing a double burden of disease; while continuing to deal with malaria and other infections, there is a rapid upsurge in NCDs [1]. This is especially alarming considering that an estimated two-thirds of diabetic individuals in the African region are undiagnosed [38]. Moreover, with the recent changes in malaria transmission in many areas, natural acquired immunity against uncomplicated and severe malaria will probably be affected [39]. Increasing prevalence of comorbidities such as obesity and diabetes might make these populations more vulnerable to severe malaria.

CONCLUSIONS
This is, to our knowledge, the first study investigating the impact of NCDs on severity of P. falciparum malaria. We show that obesity, diabetes, and combinations of metabolic risk factors were associated with severe malaria, both in nonimmune travelers and immigrants from sub-Saharan Africa. The findings are of high clinical relevance for the acute management of malaria in travelers; there is also an urgent need for awareness and further investigations of these risk factors in malaria-endemic areas.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes
Author contributions. Study concept and design: A. F., P. N., K. W. Data acquisition, analysis, or interpretation: all authors. Statistical analysis: U. F., K. W., P. N. Drafting the manuscript: K. W., A. F., M. V. Revising the manuscript critically for important intellectual content: all authors.