Background

Protein-energy malnutrition (PEM) in children constitutes a global health challenge in developing countries of sub-Saharan Africa and southern Asia [1]. Children with PEM have immunological dysregulation [2] and are thus susceptible to common childhood infections such as infectious diarrhea, pneumonia and bacteremia which, in turn, create a vicious cycle with malnutrition [3]. Similarly, these children are also thought to be mainly predisposed to urinary tract infection (UTI) as the infection risk may also increase with the severity of malnutrition [4], although there appears to be inconsistent evidence linking the degree of malnutrition to higher risk of UTI [5].

The presence of urinary secretory IgA (sIgA) is one of the defense mechanisms against UTI, and its role in UTI episodes has been reported [6,7,8]. Low urinary sIgA may represent an important predisposing factor to recurrent UTI [9]. Among other effects on the immune system, malnutrition specifically leads to diminished IgA response. A study on experimental animal models showed that dietary protein played a significant and site-specific role in the developmental expression of the secretory immune system, with severe protein malnutrition suppressing this immune arm [10]. Therefore, UTI risk in malnourished children may partly be related to impaired sIgA response.

Several studies have been conducted on UTI prevalence rates and bacterial etiologic patterns in malnourished children across the globe [11,12,13,14,15,16,17,18,19]. A 2013 systematic review of severely malnourished under-five children revealed a high prevalence of pneumonia (34%), diarrhea (35%) UTI (24%) and bacteremia (17%), with higher mortality rates compared to other children [20]. Furthermore, a more recent non-systematic review revealed vast differences in the prevalence rates with no regional disparities regarding the bacterial isolates, even though sensitivity patterns varied remarkably [5]. There was also no consensus on sex predominance of UTI among malnourished children in some of these studies [11, 12, 18], and controversy still exists on whether UTI risk in these children increases with the severity of malnutrition given the discordant reports about this correlation. Although few studies have compared UTI prevalence in malnourished vs. healthy children, there have been no pooled risk studies directly quantifying the risk of UTI due to malnutrition [21].

We, therefore, conducted a systematic review and meta-analysis to provide estimates of pooled UTI prevalence among malnourished children and of combined UTI risk in comparison with their well-nourished counterparts without age limits and including all degrees of malnutrition. These combined data should provide sufficiently robust evidence to justify the inclusion of screening and treatment of UTI in the management of children with PEM.

Methods

Search strategy and selection criteria

We systematically searched electronic databases including MEDLINE, EMBASE, Web of Science, and African Journals Online from inception till 2018 (date of the last search: 22 December 2018). We searched both databases using the following keywords alone and in combination: urinary tract infection, bacteriuria, pyuria, malnutrition, protein-energy malnutrition, severe acute malnutrition, prevalence, incidence, risk, children and infants.

Inclusion and exclusion criteria

To be included in this review, primary studies had to be observational studies of children (irrespective of origin, ethnic, socioeconomic, and educational background) reporting the prevalence of UTI with background malnutrition or with enough data to compute these estimates. We also included studies which reported an association between malnutrition and UTI or at least UTI prevalence in both malnourished and comparative healthy controls in the same research, enabling the estimation of associations. Both malnutrition and UTI had to be clearly defined in the included studies. Malnutrition had to be defined as a function of weight for age or weight for height using validated reference methods including the World health Organization (WHO)/National Center for Health Statistics (NCHS) [22], Wellcome [23], or Gomez [24] classifications or as mid-arm circumference less than 11 cm. The grade or degree of malnutrition also had to be clearly defined. When absent, we categorized grade I as mild malnutrition, grade II as moderate malnutrition, and grade III as severe malnutrition. UTI had to be defined as significant bacteriuria or pyuria corresponding to the urine sampling method. We included only full-text articles in the English language. We excluded abstracts, letters, reviews, commentaries, editorials, and studies without primary data or explicit description of methods. Two of the investigators (SNU and ICE) independently screened the titles and abstracts of articles retrieved from the literature search. Full texts of articles found potentially eligible were obtained and further assessed for final inclusion. All duplicates were removed during the study selection process. Disagreements were resolved through discussions between the investigators until a consensus was reached.

Quality assessment

We evaluated the methodological quality of included studies using the Newcastle-Ottawa Scale for assessing non-randomized studies [25]. This scale evaluates case-control and cross-sectional studies using criteria categorized into selection (4 points), comparability (2 points), and exposure/outcome (3 points). Quality Rating was categorized as low (< 7) or high (≥7). Two of the investigators (SNU and ICE) independently assessed study quality, with disagreements resolved by consensus.

Data extraction

Two of the investigators (SNU and ICE) independently extracted relevant data from individual studies using a preconceived and standardized data-extraction form. Information retrieved included the first author’s name, year of publication, year of study, study setting and country, study design, study population, sample size, and age and sex distribution of participants. We extracted information on urine sampling and analytic methods, UTI and malnutrition diagnostic criteria, the proportion of participants with UTI, and the reported population subgroup differences in proportions. We also extracted information on bacterial isolates and their antibiotic-sensitivity patterns when available. Where relevant data were not available, we contacted the corresponding author to request for the information. We assessed the inter-rater agreement for study inclusion and data extraction using Cohen’s κ coefficient [26].

Data analysis

A meta-analysis of prevalence studies

The synthesized study-specific estimates were pooled using random effects meta-regression model to obtain an overall summary estimate of the prevalence across studies, after stabilizing the variance of individual studies with the use of the Freeman-Tukey double arcsine transformation [27]. Random-effect models give more weight to smaller studies and have wider confidence intervals because they consider potential variation between the actual effects that all included studies estimate, in addition to their within-study variance. We calculated the I2 and tau2 to assess between-study heterogeneity. We assessed publication bias using funnel plots and the formal Egger [28], and Begg’s tests [29]. We considered any test p-values less than 0·05 to be indicative of significant publication bias. We assessed subgroup differences in prevalence estimates based on factors such as sex (males/females), age group (< 18 months/≥18 months), malnutrition severity (moderate or severe /mild or mixed), region of origin (Africa/Others), study design (cross-sectional/ case-control), study quality (low/high), year of study (< 2000/≥2000) and urine-collection method (one method/multiple methods or unspecified). We performed sensitivity analyses, including fixed effect meta-regression, leave-one-out random effects meta-regression to explore the stability of our pooled prevalence estimate. We assessed the sensitivity of the combined estimates to the exclusion of studies with < 30 participants, and studies where urine sampling method, urinalysis method or definition of UTI were not stated. We performed meta-regression analysis using study-level covariates as predictors of study-level estimates to explore the determinants of potential heterogeneity in our pooled estimates, in bivariate and multivariate models.

A meta-analysis of association studies

We pooled the reported or derived estimates of association (odds ratios (OR) and 95% CI) between malnutrition and UTI from included case-control studies also using random effects meta-regression model. When no UTI was reported for the control group, we added 0.5 to all four related groups to estimate the OR and 95% CI in the affected studies [30, 31]. We also calculated the I2 and tau2 to assess between-study heterogeneity and evaluated publication bias using funnel plots and the formal Egger [28], and Begg’s tests [29]. We performed subgroup analysis, including stratification by matching (i.e., if the studies matched the Cases and Controls by at least age or sex, or both). We also performed sensitivity analyses, including fixed-effect and leave-one-out meta-regression. Given the limited number of association studies (n = 8), we could not perform further meta-regression analyses. Data were analyzed using STATA version 14.0 for Windows (STATA Corporation, Texas).

For reporting, we adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis [32], and the Meta-analysis Of Observational Studies in Epidemiology guidelines [33]. This systematic review is registered with PROSPERO, number- CRD42018084765.

Results

Study selection

We identified 1478 records following a combined search of MEDLINE, EMBASE, ISI Web of Science, and African Journals Online databases. Exclusion of duplicates and non-pertinent articles yielded 35 articles, of which 33 met the eligibility criteria. A further search of the references of these articles yielded an additional item. Thus, the present review includes 34 full-text articles, either reporting UTI in malnourished children only or parallel with well-nourished children. Details of the article-selection algorithm are presented in Fig. 1.

Fig. 1
figure 1

Algorithm for inclusion in the present study

Full size image

Characteristics of included articles

Overall, we included 26 cross-sectional (76%) and 8 case-control studies (24%). All included studies were hospital-based studies. Most of the studies were from African countries including South Africa [13,14,15, 34,35,36] Nigeria [11, 37,38,39,40], Uganda [41, 42], Kenya [17, 43] Tanzania [18, 44], Ethiopia [16] Niger [12], Sudan [45], and Gambia [19]. Other studies were conducted in Turkey [46, 47], India [4, 48, 49], Pakistan [50, 51], Bangladesh [52], Thailand [53], Iran [54], Australia [55], Peru [56], and Jamaica [57]. Sample size varied, with 18% of the studies having < 50 participants. Eight cross-sectional studies (31%) primarily investigated UTI in malnourished children [13, 14, 17, 18, 35, 46, 53, 57], whereas the remaining 18 studies (69%) reported UTI as a secondary outcome in the broader context of bacterial infections in malnourished children [11, 12, 15, 16, 19, 37,38,39,40,41,42,43,44,45, 50,51,52, 55]. Most of the case-control studies (88%) primarily investigated UTI occurrence in malnourished children vs. healthy controls [4, 34, 47,48,49, 54, 56], whereas one (12%) reported UTI prevalence in both groups in the broader context of bacterial infections in children [36]. The pooled study population included 3294 malnourished children from 26 cross-sectional studies and 2051 children (1052 malnourished and 999 controls) from the 8 case-control studies, for estimating the pooled prevalence of UTI and pooled OR and 95%CI of UTI with malnutrition, respectively (Table 1).

Table 1 Characteristics of studies on malnutrition and urinary tract infection
Full size table

Most of the studies included participants who had moderate-to-severe malnutrition (76%), while the rest had mixed malnourished populations (24%). There were differences in urine-sampling methods with most studies employing two or multiple methods (74%) including combinations of suprapubic aspiration, mid-stream urine or urine bags [4, 16, 18, 34,35,36, 38,39,40, 48, 49, 52, 53, 57], compared to a single method (26%) in their study population [11,12,13, 17, 43, 47, 54, 56] (Table 2). There was uniformity in the definition of UTI across studies, which was consistently applied to the urine-sampling method (Table 2). Although all studies examined the prevalence of UTI in malnourished children, 38% of the included studies did not explicitly describe urine-collection method, urine analytic method, or UTI (Table 2).

Table 2 Definition of malnutrition and urinary tract infections across included studies
Full size table

UTI prevalence in malnourished children

As shown in Fig. 2, the pooled random-effects prevalence of UTI in 3294 malnourished children was 17% (95% CI: 13, 21%). Heterogeneity was high across studies (I2 = 87.6%; P < 0.001; Tau2 = 0.06). Subgroup analyses showed significant differences by degree of malnutrition (severe: 15% (95% CI: 11, 19%); mild/mixed: 25% (95% CI: 19, 32%); Pheterogeneity: 0.01) and sample size (Sample size < 50: 27% (95% CI: 18, 36%); Sample size ≥50: 16% (95% CI: 12, 20%); Pheterogeneity: 0.02), and borderline-significant differences by year of study (year < 2000: 21% (95% CI:16,26%); year≥2000: 14% (95% CI: 9, 19%); Pheterogeneity: 0.06). We did not observe significant differences by age group (Pheterogeneity: 0.21), study region (Pheterogeneity: 0.68) and study quality (Pheterogeneity: 0.33). Although the difference by urine sampling method was non-significant (Pheterogeneity: 0.29), the prevalence of UTI in studies which applied suprapubic aspiration or sterile catheterization alone was 14% (95% CI: 7, 22%) while that of those combining different methods was 18% (95% CI: 14, 23%). Sex-specific prevalence of UTI in malnourished children was similar among the six studies reporting these estimates (UTI prevalence in males: 23% (95% CI: 14, 32%); females: 20% (95% CI: 14, 27%); Pheterogeneity = 0.61) (Table 3). Figure 3 shows the funnel plot for visualization of publication bias. We observed minimal evidence for publication bias as both Egger’s (P = 0.15) and Begg’s tests (P = 0.35) were non-significant. Further sensitivity analyses revealed the robustness of our findings. Fixed-effects prevalence of UTI in malnourished children was 15% (95% CI: 14, 17%) (Fig. 2) whereas exclusion of studies with <30 participants or not specifying urine sampling or analytic method or UTI definition yielded a random-effects pooled prevalence of 17% (95% CI: 13, 21%) and 20% (95% CI, 14, 27%) respectively (Additional file 1: Table S2).

Fig. 2
figure 2

Overall UTI prevalence in malnourished children

Full size image
Table 3 Subgroup random-effects prevalence estimates of urinary tract infection in malnourished children
Full size table
Fig. 3
figure 3

Funnel plot for visualization of publication bias with studies reporting UTI prevalence

Full size image

Results from meta-regression analyses including study-level covariates showed the degree of malnutrition, sample size and year of study to be significant predictors of prevalence rates, explaining 24.1, 5.8 and 5.7% of the between-study variance respectively in the bivariate models, respectively. Degree of malnutrition remained significant in the multivariate meta-regression model that also included sample size and year of study. Studies, including severely-malnourished children, reported a lower prevalence of UTI compared to a milder/mixed group (OR: 0.90 (95% CI: 0.83, 0.97)). Although statistically non-significant, prevalence of UTI also decreased with sample size (OR: 0.92 (95% CI: 0.83, 1.02)) and studies published from 2000 (OR: 0.95 (95% CI: 0.89, 1.02)). This multivariate meta-regression model explained 33.9% of the between-study variance in the pooled estimates (Table 4).

Table 4 Meta-regression estimates to explain the prevalence of urinary tract infection in malnourished children
Full size table

Risk of UTI in malnourished children vs. healthy controls

Random-effects pooled OR of UTI in 1052 malnourished children, and 999 controls were 2.80 (95% CI: 1.41, 5.54). We observed moderate heterogeneity in across studies (I2 = 53.6%; P = 0.04; Tau2 = 0.47) (Fig. 4). Stratifying by matching criterion showed differences in random effects associations between UTI and malnutrition (OR matched studies: 5.67 (1.39, 23.2); I2 = 56.7%; P = 0.07; Tau2 = 1.09; OR in unmatched studies: 2.04 (0.91, 4.57); I2 = 57.4%; P = 0.07; Tau2 = 0.38). Figure 5 shows the funnel plot for visual assessment of publication bias within the case-control studies. We also observed minimal evidence for publication bias given the non-significant Egger’s (P = 0.34) and Begg’s tests (P = 0.90). Sensitivity analyses revealed robust effect estimates. Fixed effect pooled OR of UTI was 2.50 (95% CI: 1.66, 3.89) (Fig. 3). Leave-one-out random effects OR of UTI ranged from 2.34 (1.19, 4.62) to 3.26 (1.63, 6.50). We observed the smallest heterogeneity (I2 = 47.2%; P = 0.08; Tau2 = 0.41) on the exclusion of the study by Banapurmath and Jayamony [48].

Fig. 4
figure 4

Meta-analysis of overall UTI prevalence rate

Full size image
Fig. 5
figure 5

Funnel plot for visualization of publication bias with the case-control studies

Full size image

Bacterial isolates and antibiotic-sensitivity patterns

Urine culture was performed by 28 (82%) of the included studies. Of the 27 studies that reported urinary bacterial isolates, Escherichia coli was the predominant isolate in 25 (93%) of them, whereas Klebsiella spp. was predominant in 2 (7%). Most common bacterial strains included gram negative coliforms, including Escherichia coli (100%), Klebsiella spp. (81%), Proteus spp. (41%), Pseudomonas spp. (33%), Enterobacter spp. (22%), and Citrobacter spp. (15%). Other reported gram-negative bacterial isolates include Salmonella spp. (7%), Serratia spp. (7%), Hafnia alvei (4%) and Morganella morganii (4%). Gram-positive isolates were less prevalent and included Staphylococcus spp. (7%), Enterococcus spp. (7%), and Streptococcus faecalis (4%) as well as the fungus, Candida albicans (4%). Antibiotic sensitivity tests were performed by 13 (38%) studies, with different sensitivity patterns (Table 5).

Table 5 Prevalence of urinary tract infections (UTI) and bacterial isolates in malnourished children across included studies
Full size table

Comorbidities of UTI in malnourished children

The most commonly reported morbidities in malnourished children were diarrhea or gastroenteritis (53%; n = 18) [12, 15, 16, 37,38,39, 41, 42, 44,45,46, 48, 50, 52,53,54,55,56], respiratory diseases (including pneumonia, tuberculosis, respiratory tract infection and abnormal chest radiographs; 47%; n = 16) [12, 16, 19, 37,38,39, 41, 42, 45, 46, 48, 50, 52,53,54,55] and bacteremia or sepsis (47%; n = 16) [12, 14,15,16,17, 19, 37,38,39, 41,42,43,44, 50, 52, 53]. Six studies reported co-occurrence of UTI with at least one of these common malnutrition-associated morbidities [12, 14, 19, 39, 48, 54]. Only 27% (n = 7) of the cross-sectional studies on UTI in malnourished patients investigated renal urinary tract malformations in their UTI patients [4, 36, 48, 49, 54, 56], reporting a combined malformation prevalence of 14% in these patients. In contrast, 75% (n = 6) of the case-control studies utilized radiological investigations to identify malformations as a risk factor for UTI in their patients, reporting a prevalence of 34% (n = 80) among the malnourished children and a prevalence of 4% (n = 4) among the healthy controls.

Discussion

This paper is the first PROSPERO-registered systematic review on UTI among malnourished children. In this review and meta-analysis of data from 34 studies involving 3294 malnourished children, we found a pooled UTI prevalence of 17% and pooled OR of 2.34 for UTI in association with malnutrition in 2051 children (1052 malnourished children versus 999 controls). Our combined prevalence rate is at variance with the rate of 24.1% reported in a systematic review on the justification for antibiotic use in children with uncomplicated severe acute malnutrition (SAM) [20]. The disparity could be due to differences in the number of reviewed studies (26 in the current study versus 10 in the comparative study), and may also be explained by the predominant age bracket of the malnourished children reviewed by these authors [20], which fell within the period of pre-toilet/toilet training: a phase that contributes to UTI risk in childhood [21]. The systematic review by Alcoba et al. specifically selected studies that investigated the prevalence of other infections, such as human immunodeficiency virus, bacteremia, lower respiratory tract infection, and diarrhea in strictly SAM and not-only-SAM children [20]. However, the prevalence rate from our review is similar to the 11–16.5% prevalence reported in the selected studies from the West African sub-region [11, 12, 19, 38, 39], India [4, 49], Turkey [47], and Australia [55].

We found no significant sex predominance in the few studies that reported a sex-specific prevalence of UTI in malnutrition. This finding is inconsistent with the known epidemiologic trajectory of UTI in which prevalence rates for both sexes may be the same during infancy, but show male predominance in the neonatal period and female preponderance during early childhood and the period of toilet training [5]. More importantly, the later female dominance may be due to anatomical differences where the proximity of the urethral opening to the vagina may facilitate urethral contamination [58]. In addition, recent evidence suggests that the sex differences in the reticuloendothelial system which provides innate immunity against microbes may also contribute to the sex differences in UTI prevalence rates [59]. Thus, irrespective of nutritional status, female sex remains a risk factor for UTI in childhood. We also noted that UTI risk was increased by the severity of malnutrition. Its prevalence was slightly higher in children aged less than 18 months. Although the latter observation may be related to exposure to gut uropathogenic bacterial flora during the period of pre-toilet training, the former agrees with the report of one of the selected studies which showed a direct correlation of UTI risk with the severity of malnutrition [4]. It is however in contrast with the findings of studies in Nigeria [11], and South Africa [14], which did not establish any significant change in UTI prevalence rates for the different grades of malnutrition. The lower prevalence of UTI in the severely malnourished children may be related to their lower efficiency in immune response due to lack of immune cells and immune dysfunction which characterize severe malnutrition [60]. Although non-significant, the higher prevalence of UTI in studies combining several sampling techniques (that included less sterile methods) might have been due to contamination in the collection process. But the allowance of up to 105 colonies per ml in the diagnosis of UTI (when the reference method of suprapubic aspiration is not used) limits outcome misclassification, and could explain the non-significant difference observed in our study.

Our finding of a positive and significant pooled risk of UTI in malnourished children compared to healthy control is not surprising given their higher susceptibility to infections based on their immune dysregulation. We also found a consistent report of higher occurrence of other infections across studies which investigated other concurrent infections. Malnourished children also had higher prevalence of urinary tract anomalies, which is a known risk factor for UTI [61].

Another key finding in our systematic review is the predominance of Escherichia coli and other gram-negative coliforms as the bacterial isolates. This trend is similar in both malnourished and non-malnourished children. It is trite to mention that exposure of children to infection with gut uropathogens (during pre-toilet and toilet-training periods) is a putative UTI risk factor, which may partly explain this observation. Apart from the role of malnutrition in causing diminished IgA response (including sIgA), the reduced transferrin levels in malnourished children may result in the circulation of free unbound iron, which creates a favorable environment for the growth of gram-negative bacteria leading to gram-negative sepsis and subsequently UTI via the hematogenous route [5].

There are substantial differences in the antibiotic-sensitivity patterns of the predominantly isolated gram-negative bacteria, including Escherichia coli. Our observation across the reviewed studies clearly shows no defined pattern of sensitivity and resistance to the tested antibiotics. This finding underscores the need for a periodic institution-based update of antibiotic-sensitivity trends. Relying on previous sensitivity reports as guides for empirical therapy may result in poor outcomes for new cases of UTI in malnourished children.

The strengths of our study include its broad approach in identifying relevant articles, and the consideration of both UTI prevalence in malnourished children, and the risk of UTI in malnourished children vs. controls. We explored publication bias, and the determinants of high heterogeneity observed in our estimates. Our inclusion of a large number of studies also allowed for sensitivity analyses, which confirmed the robustness of our pooled estimates. However, our research has some limitations. First, we observed high heterogeneity across the studies included in the combined prevalence estimates. While we identified some factors that explained some of the between-study heterogeneity, other unmeasured factors could have also contributed as we could only explain 34.6% of this heterogeneity. The inclusion of earlier studies may have biased our pooled estimates given the continuous updates of definitions and management protocols for childhood diseases. However, the definition of UTI and other methodologies were quite similar across included studies, and although stratification by year of publication showed higher prevalence of UTI among older studies (before 2000; pooled prevalence of 21%), the prevalence of newer studies (2000 and later; pooled prevalence of 14%) was similar to the overall pooled random effects estimate (17%). Year of publication was also not a significant determinant of between-study heterogeneity in the meta-regression model (Table 4). Our observation of the absence of publication bias in the pooled OR of UTI might have been due to the small number of studies, as there is a high risk of non-detection of publication bias in meta-analyses that include less than ten publications [62].

In conclusion, our systematic review has shown that UTI is more prevalent in malnourished children than in their well-nourished counterparts. It has been suggested that if children at high risk of UTI like those with malnutrition were screened, the number of children missed or treated inappropriately could be reduced [63]. We recommend the incorporation of screening and treatment for UTI into the management protocol for malnourished children to improve disease outcomes.