Association between nutritional status and dengue infection: a systematic review and meta-analysis

In August 2013, we searched for eligible studies from eleven electronic databases. PubMed and Information Sciences Institute (ISI) were searched by using the following query: dengue AND (nutrition OR nutritional OR malnutrition OR malnutritional OR “body mass index” OR obesity OR overweight). In Scopus, we used the search terms “TITLE-ABS-KEY(dengue AND (nutrition OR nutritional OR malnutrition OR malnutritional OR "body mass index" OR obesity OR overweight))”. In Google Scholar, studies were retrieved by this query: “allintitle: dengue (nutritional OR "malnutrition" OR "body mass index" OR "obesity" OR "overweight")”. We performed the search terms “dengue AND (nutritional OR "malnutrition" OR "body mass index" OR "obesity" OR "overweight")” on WHO Global Health Library and Popline. As for Indexing of Indian Medical Journals (IndMED), the following search terms were used: “dengue AND ((nutritional) OR (malnutrition) OR (BMI) OR (obesity) OR (overweight))”. In addition, we searched the keyword “dengue” on the remaining libraries including African Journals OnLine (AJOL), African Index Medicus (AIM), New York Academy of Medicine Grey Literature Report (NYAM) and System for Information on Grey Literature in Europe (SIGLE). Moreover, we manually collected studies by screening the references of relevant reviews and included studies [16‐20].

All titles and abstracts (when available) were reviewed independently by at least two of seven authors (NTHT, NPL,TTMH, TDT, DND, LPH) after a pilot training of 20 % relevant articles with a senior researcher (NTH). Studies were considered eligible if they stated any information of dengue infected patients together with information of nutritional status. There were no restrictions with respect to publication language, patient age (children or adult) or study design. Non-English reports were translated into English by authors with the help of native international students at Nagasaki University. We excluded articles with the following characteristics: (i) including data that could not be reliably extracted; (ii) including data sets considered overlapping; (iii) letter, case report, review, thesis or conference paper; (iv) animal study or in vitro study without patients. Any conflicts were resolved by discussion and consensus between authors.

Full-text versions of all eligible studies were obtained. Data were extracted by two independent reviewers (NTHT and NPL) and were checked by at least two of four authors (NTH, NTHT, NPL, TTMH). Any disagreement was resolved by discussion and consensus. The data extracted included the first author, year of publication, year of patient recruitment, study design (cross-section or case-control), data collection (prospective or retrospective), assignment of patients (consecutive or random), country and city of origin, hospital where the patients were recruited, characteristics of the patient population (infant, children or adult), number of included individuals, criteria of dengue infection (confirmed or clinical diagnosis), criteria of DSS, DHF and DF, data about nutritional status (malnutrition, normal or obesity) of included individuals. We also recorded a description of blinded interpretation of factors, the age of the patients and Body Mass Index (BMI).

Fifteen of thirty-three studies were included for extracting data at the first time. However, two studies did not have suitable data to perform meta-analysis [4, 21]. We contacted twice via email these authors to get more information but there was no reply. Consequently, thirteen studies were incorporated into final analysis.

Quality assessment was independently carried out by two investigators (NTHT, NPL). The quality of included studies was assessed based on the combined criteria suggested by Pai et al. [22] and Wells et al. [23]. The quality of each study was determined across eight metrics: study design, full description of characteristics of patient population (infant, children and adult), data collection (prospective or retrospective), assignment of the patients (consecutive or random), inclusion criteria, exclusion criteria, blinded interpretation of factors, and full description of dengue diagnosis. Quality was evaluated by discussion and consensus after the independent review of each study by four authors (NTH, NTHT, NPL, TTMH) (Additional file 2: Table S2).

Meta-analyses including sensitivity, subgroup, meta-regression, and publication bias analyses was performed using Comprehensive Meta-analysis software version 2 (Biostat, USA, https://​www.​meta-analysis.​com/​) as previously described [14].

After filtering and deleting the duplicates by Endnote software, a total of 362 publications were included for initial screening of titles and abstracts (Fig. 1). Full-text reading was continued in 33 studies that met the inclusion and exclusion criteria. We excluded 25 articles with the following reasons: (1) study without relevant items; (2) content not satisfying criteria; (3) overlapping data; (4) unable to extract data; (5) duplicated reports. Finally, 13 reports (8 studies from electronic databases and 5 studies retrieved from manual search) came into the final analysis.

Most of the included studies were performed in Asia (5 from Indonesia, 4 from Thailand, 2 from Viet Nam and 1 from India) while only one study was from El Salvador, a Latin American country [7, 8, 10, 12, 13, 16‐20, 24‐26]. The characteristics of these studies are outlined in Additional file 3: Table S3. We identified eight cross-sectional studies and five case-control studies. Ten studies were prospective and the remaining were retrospective or not mentioned (n = 3). The subjects of 13 selected researches were both children and/or infants. Serological diagnosis (ELISA or hemagglutination inhibition), PCR and viral isolation were some laboratory diagnosis methods used in dengue infection. All the thirteen studies used the WHO classification of dengue severity in which nine studies used the WHO 1997 criteria, two used WHO 1986, one used WHO 1999 and one study did not mention the particular year of WHO classification.

The definition of malnutritional status was different between studies. Nine studies used weight-for-age to assess malnutritional status [8, 10, 12, 13, 16, 18, 24‐26]. Two of these nine studies [10, 25] used both weight-for-age and height-for-age while other three [7, 19, 20] did not give a clear method to evaluate the malnutrition status. In addition, three studies also used weight-for-height [10, 17] and BMI-for-age [25] to measure this factor.

We analyzed 10 studies in which the authors made the comparison between DSS group and DHF group to evaluate the correlation between malnutrition and suffering DSS [7, 8, 10, 12, 13, 16‐19, 26] (Fig. 2a). Pooling of these 10 studies revealed no significant correlation (OR = 1.17, 95 % CI: 0.99–1.39) (Table 1). Only the study of Kalayanarooj et al. showed a positive relation (OR = 1.42, 95 % CI: 1.12–1.80). Subgroup analysis of seven studies using nutrition assessment method by weight-for-age index showed that malnutrition status was related to DSS (OR = 1.25, 95 % CI: 1.02–1.53) [8, 10, 12, 13, 16, 18, 26] (Fig. 3a). However, removing the largest study of Kalayanarooj et al. [8] resulted in the loss of this correlation (p = 0.5). Three of the ten studies mentioned above permitted us to compare the ratio of malnourished children between DSS group and DHF plus DF group [7, 8, 19]. No significant correlation was established (OR = 1.21, 95 % CI: 0.99–1.48, Fig. 3e). When conducting a comparison between DSS plus DHF group to DF group by pooling these three studies, we detected an inverse correlation of malnutrition with developing hemorrhagic tendency (OR = 0.77, 95 % CI: 0.61–0.97). It is almost the same as the result of the study of Kalayanarooj et al. (OR = 0.77, 95 % CI: 0.61–0.98, Fig. 3b).

There were five studies evaluating DHF group and DF group to identify the relationship of malnutrition with developing DHF [7, 8, 19, 24, 25]. There was a significantly inverse relation (OR = 0.71, 95 % CI: 0.56–0.90). Among the five studies, only the study of Kalayanarooj et al. showed a significant relationship (Fig. 3c) and the pooled result became non-significant after removing the study of Kalayanarooj et al. [8] (data not shown).

Finally, meta-analyzing three studies [10, 13, 25] showed a statistically significant negative correlation of malnourished children with dengue virus infection with the healthy children (OR = 0.46, 95 % CI: 0.3–0.70, Fig. 3d, Additional file 4: Table S4). Both studies of Hung et al. and Thisyakorn et al. revealed negative correlations. However, these two studies compared DSS plus DHF group with healthy group [10, 13], while other compared DHF plus DF group with healthy group [25].

Pooling nine studies [8, 10, 12, 16‐20, 26], all of which were performed in South-East Asia, indicated that children with normal nutrition were inversely associated with DSS compared to DHF (OR: 0.87, 95 % CI: 0.77–0.99, Fig. 2b). Neither evidence of heterogeneity (p = 0.26, I ²  = 21) nor publication bias (p = 0.43) existed. None of the nine studies revealed a significantly positive association of normal nutrition with DSS while one study showed a clearly negative relation of this factor with DSS [17]. Moreover, further analysis of eight studies after excluding any of three studies [8, 17, 20] did not provide a significant association of normal nutrition with DSS.

When comparing group of DSS with group of DF/DHF (Fig. 4a), no correlation between normal nutrition and DSS was found (OR: 0.93, 95 % CI: 0.8–1.07) after pooling from two studies [8, 19]. Similarly, when comparing group of DSS/DHF with group of DF, the analysis did not find a statistically significant relation of normal nutrition with developing hemorrhagic tendency or shock (OR: 1.06, 95 % CI: 0.91–1.24, Fig. 4b). Moreover, a meta-analysis of four studies [8, 19, 24, 25] showed no significant association with normal nutrition when comparing DHF with DF (OR: 1.06, 95%CI: 0.90–1.23, Fig. 4c).

Pooled result from eight studies [8, 11, 12, 16‐19, 26] revealed no significant association of overweight/obesity with DSS compared with DHF (OR: 1.31, 95 % CI: 0.91–1.88, Fig. 2c), although two studies showed positive correlation [11, 17]. There was no evidence of publication bias (p = 0.32). Further subgroup analysis of six studies which used weight-for-age index did not find any statistically significant correlation between two factors mentioned above (OR: 1.01, 95 % CI: 0.87–1.19, Fig. 5a) [8, 11, 12, 16, 18, 26].

When we compared group of DSS with group of both DHF and DF, no correlation of overweight/obesity with suffering DSS was found (OR: 0.99, 95 % CI: 0.84–1.16, Fig. 5b). We got the same result when comparing group including DSS and DHF with group of DF (OR: 1.05, 95 % CI: 0.88–1.25, Fig. 5c). Besides the study of Basuki et al. and the study of Kalayanarooj et al., there are two additional studies in which the authors compared between DF and DHF patients [24, 25]. The pooled result showed that overweight/obesity was not found associated with the dengue severity (OR: 1.06, 95 % CI: 0.89–1.26, Fig. 5d).

All data is presented within the manuscript and its supplementary files.

Not applicable.

Not applicable.