Search strategy

This review, with pertinent information regarding our review protocol, was prospectively registered with the PROSPERO database as follows: http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42016040052 (see Supporting Information file S1 Text). We searched Medline, Web of Science, and Biosis electronic databases for original research articles, conference abstracts, or dissertations available as of August 22, 2016. Databases were searched with pre-specified keywords including “bilharzia”, “schistosomiasis”, “Schistosoma”, “school attendance”, “attention”, “impairment”, “memory”, “cognition”, among others. The complete search strategy is detailed in S2 Text.

Population, inclusion and exclusion criteria, study design

This systematic review and meta-analysis is focused on school-aged children five years and older. We did not restrict studies according to language, design, or publication date. Both interventional and observational studies were included in this review if they evaluated cognitive function in school-aged children using any psychometric test, or measured school attendance or achievement in relation to infection by Schistosoma parasites of any species. We excluded studies exclusively focused on pre-school aged children because of the absence of educational measures and the use of neurocognitive tests (e.g., the Mullen test) that were difficult to classify in terms of neurocognitive domains. We excluded meta-analyses and primary studies of soil-transmitted helminth infections where Schistosoma coinfection was absent.

Comparisons: Schistosoma infection vs. no infection & treatment with praziquantel vs. placebo

Schistosoma infection status was determined by microscopic examination of stool or urine as appropriate for species. Praziquantel was the primary deworming agent in interventional studies. The primary exposure for this meta-analysis included presence of infection or, operationally, infection was categorically defined based on study design as follows: 1) untreated/placebo versus praziquantel-treated in a randomized controlled trial, 2) any, versus no Schistosoma infection in cross-sectional studies, or 3) pre-, versus post-praziquantel treatment, or infection-free versus persistent infection, among Schistosoma-infected individuals in a longitudinal design study. “Untreated” refers to children determined to be infected but not dewormed.

Outcomes: Psychometrically assessed cognitive function

In the reviewed studies, psychometrically-assessed cognitive function was measured by a range of instruments that, for the purposes of this meta-analysis, were categorized into four domains: memory, learning/executive function, attention/reaction time, and intelligence (see S1 Table for details). The memory domain instruments included tests of working (short-term) memory as well as those of long-term memory. Attention/reaction time tests were those that measured the ability of a child to sustain concentration on a particular object, action, or thought, including their capacity to manage competing demands in their environments. The learning/executive function domain included tests to evaluate children’s performance in goal-oriented behavior, particularly components that are important for scholastic advancement. These test a cluster of cognitive processes that enable children to connect past experience with present action, and by so doing, engage in planning, organizing, strategizing, paying of attention to details, and to emotionally self-regulate, make necessary efforts to remember important details required for attainment of future goals [29]. We included in the ‘intelligence’ domain psychometric tests of intelligence quotient (IQ) that likely measures largely biologically-determined cognitive abilities, in contrast to cognitive performance measures that are environmentally pliable [33].

It was common for studies to use a suite of psychometric instruments to assess a single or multiple cognitive domains in enrolled children. When multiple instruments were used to measure the same cognitive domain, a grand mean of scores and a grand mean of standard deviation (SD) across all instruments were calculated. Thus, for each publication, one overall mean and SD value was determined for each domain. A study could contribute data to different cognitive domains if it used tools spanning across several cognitive domains. However, each instrument only contributed to one single domain of function, as shown in S1 Table.

Outcomes: Educational loss

Two dimensions of educational loss were tabulated–school attendance and scholastic achievement for children enrolled in school.

School attendance: For children enrolled in cross-sectional and longitudinal studies, attendance rate was respectively defined as the number of days children attended school over the past month or over the study period. In case-control studies, the percentage of children enrolled vs. not enrolled in school was calculated for Schistosoma-infected and non-infected children.

School achievement was assessed across studies based on: i) children’s pass rate on standardized teacher-generated tests; ii) the percent of children who were in appropriate class for age; iii) their enrollment in elite vs. non-elite schools; iv) their scores in the school function domain of pediatric quality-of-life inventory; v) their change in class position after treatment for Schistosoma infection; vi) an above average vs. average/below average scholastic performance as rated by a teacher; or vii) their pass rate in any kind of educational test, whether teacher-administered or not.

Study selection, data extraction and management

Two researchers (LM and AK) independently screened individual articles by title and abstract, after which eighty-eight full text articles were assessed for eligibility for inclusion in this review. Studies were excluded on the following basis: no outcome measure reported (n = 39), non-primary literature or a review article (n = 7), absence of both infection and outcome measures (n = 6), no variation in exposure (n = 2), a limited meeting abstract duplicated by later full publication (n = 3) and a nonhuman study (n = 1) Disagreements between reviewers on inclusion of a given study were resolved by consensus between LM and AK. If no consensus was reached, the article was further evaluated by an additional reviewer (AEE). Thereafter, two researchers (AEE and NP) independently extracted relevant data for meta-analyses. Where differences in approach to standard error (SE) estimation were noted, discrepancy was resolved by consensus between AEE and NP. When a potentially relevant publication did not present needed information for meta-analysis, the authors were contacted to request additional data. If the dataset was publicly available, we obtained needed values directly [28].

Statistical analysis

The method for deriving SD from respective studies depended on how data were presented in the original reports. Some papers presented median (m) and range (a to b) instead of means and SD. These measures were converted into approximate mean and SD as follows: , see [34], where and S² refer to the values of mean and variance, respectively. Some studies reported mean of respective measures and 95% confidence intervals. For such studies, SDs were derived as follows: SD = sqrt(N)*(upper limit–lower limit)/3.92. Other studies presented data on means and their standard errors (SE), and SD was estimated as SD = SE*the square root of N, the study size. For studies presenting data on differences in mean scores between two time points for treated/infected vs. untreated/uninfected groups, appropriate SD for mean difference was calculated using the approach recommended by the Cochrane Collaboration [35].

For the meta-analysis, studies were grouped into six categories of outcomes used to measure cognitive or school-based function: school attendance, school achievement, memory, learning, IQ, or attention. Standardized mean difference (SMD) estimates and 95% confidence intervals (CI) were calculated for each test. SMD estimates were classified as robustly statistically different if their confidence intervals excluded zero. SMD estimates were interpreted based on thresholds described by Cohen [36], as follows: ‘trivial’ (< |0.20|), ‘small’ (≥| 0.20| to < |0.50|), ‘moderate’ (≥ |50| to < |0.80|) or ‘large’ (≥ |0.80|) effects, according to standard practice in social science research.

All analyses and plots were implemented in STATA, versions 11 or 12. Heterogeneity between studies was measured with Higgins’s and Thompson’s I² statistic and chi-square p-values [37]. Where between-study heterogeneity was high, random effects modelling was used to estimate a pooled summary effect across studies [38, 39]. In the absence of heterogeneity, fixed effects modeling was performed [39]. Publication bias was assessed using the Egger test [40]. In sensitivity analysis, we evaluated potential heterogeneity in pooled impact estimates based on: i) observational vs. intervention study design; ii) the quality of original studies based on the Newcastle-Ottawa quality assessment scale; and iii) by Schistosoma species. For sensitivity analyses by species, we distinguished between urogenital schistosomiasis (S. haematobium), which is often obvious to affected children, and intestinal/hepatosplenic schistosomiasis (S. mansoni/S. japonicum). The latter two infections are similarly diagnosed by stool exam and infection is seldom obvious to most children. We examined the potential for overly influential publications using the ‘metabias’ function in STATA to evaluate robustness of our pooled estimate, based on sequential removal of individual publications from the calculation of summary estimates. Lastly, we evaluated the impact of year of publication on the stability of the pooled estimate by iteratively including studies based on year of publication–i.e. starting from the earliest to the most recent publication using the ‘metacum’ function in STATA.

Quality assessment

Our investigation was guided by recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) initiative and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines for observational studies [41]. Quality ranking of each study was implemented using an adapted version of the Newcastle-Ottawa quality assessment scale (NOQAS) [42, 43] to derive a quality score for each investigation with respect to: i) representativeness of the infected population sample or the selection cases and controls (this yielded a score with range 0 to 3*); ii) comparability with respect to known correlates of cognitive function/educational attainment (score range 0 to 6*); iii) the absence of bias in relation to outcome assessment in prospective cohort studies (0 to 3*) or exposure assessment in cross-sectional and case-control studies (0 to 3*). We adapted the comparability segment of this scale to account for confounding effects of age, sex, nutrition, and socioeconomic status in the relation between Schistosoma infection and educational/cognitive outcomes. Comparability with respect to these factors was either achieved by design (i.e. age or sex restriction for observational studies, or randomization for RCTs), or analytically, via stratified analyses or multivariable adjustment in regression models. Scores were assigned for attainment of comparability with respect to these factors as follows: age (score = 1*), sex (score = 1*), nutritional status (score = 2*) and socioeconomic status (score = 2*). For each study, the initial raw quality score (max = 12*) was rescaled to match the scale of 9* and then classified as low, high, or very high risk of bias based on precedents in prior literature [42].

Schistosoma infection and its associations with school attendance and educational attainment

Sixteen studies evaluated Schistosoma infection-related differences in school attendance (Table 2). Schistosoma infection-associated attendance deficits varied in magnitude and direction by study design. Specifically, we did not find any evidence of association between Schistosoma infection and school attendance among the two interventional studies (SMD = 0.03, 95%CI: -0.73, 0.78; Table 4); however, the observed infection-associated deficit in school-attendance was robust for the pooled estimate of 14 observational studies (SMD = -0.42, 95%CI: -0.70, -0.14). Compared to uninfected or praziquantel-treated children, the magnitude and direction of infection-associated deficit in school attendance was similar for children infected with S. haematobium or S. mansoni. Within strata of study quality, the association between infection and scholastic achievement was directionally consistent and statistically robust (Table 4). Across all studies, regardless of design or ROB, a deficit in school-attendance was evident for Schistosoma-infected or non-praziquantel-treated children compared to uninfected or praziquantel treated children (n = 15 studies; SMD = -0.36, 95% CI: -0.60, -0.12).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Pooled estimate of Schistosoma infection or non-treatment on educational and cognitive loss in school-aged children from schistosomiasis-endemic regions: Stratified by study design, Schistosoma species, and study quality^a,b.

https://doi.org/10.1371/journal.pntd.0005524.t004

The overall finding of schistosomiasis-associated achievement deficit (n = 16 studies; SMD = -0.58, 95%CI: -0.96, -0.20) was directionally consistent but varied in magnitude by study design and study quality (Tables 3 and 4). Specifically, achievement deficit with Schistosoma infection was noted in the pooled estimate derived from four intervention studies, but the magnitude of this association was lower and not statistically robust (SMD = -0.35, 95%CI: -0.71, 0.01; Table 4). Compared to uninfected or PZQ-treated children, the infection or non-treatment associated deficit in scholastic attainment was directionally consistent across different Schistosoma species, although the magnitude of effect was higher for infection with S. haematobium (SMD = -0.62; 95% CI:-1.09, -0.14) than for infection with S. mansoni (SMD = -0.22; 95% CI:-0.40, -0.05). Among observational study designs, scholastic achievement deficit was statistically robust and of larger magnitude (n = 12 studies; SMD = -0.65, 95% CI: -1.12, -0.17, Table 4). Similarly, Schistosoma infection was not associated with deficits in scholastic achievement among studies identified as low risk of bias (n = 6 studies; SMD = -0.08, 95%CI: -0.21, 0.02; Table 4). The estimates of infection-related deficit in scholastic achievement increased when study quality was lower: for studies with high risk of bias (n = 7) the SMD = -0.84 (95% CI:-1.52, -0.16); for studies with very high risk of bias (n = 7) SMD = -0.92 95% CI: -2.1, 0.28) although the estimated difference was statistically imprecise (i.e., with a wider CI) among the studies with greatest risk of bias.

Impact of Schistosoma infection on psychometrically evaluated cognitive domains

We found Schistosoma infection-associated deficits in memory tests (n = 8 studies; SMD = -0.28, 95% CI: -0.52, -0.04; Table 2). Similarly, Schistosoma infection was associated with small-to-moderate deficits in learning tests (n = 6 studies; SMD = -0.39, 95%CI: -0.70, -0.09; Table 2). The Schistosoma infection-related deficits in memory tests were directionally consistent by study design; although separate pooled estimates for interventional studies (n = 4 studies; SMD = -0.36, 95% CI: -0.61, 0.09) and observational studies (n = 4 studies; SMD = -0.15, 95% CI: -0.34, 0.03) were not statistically robust. For Schistosoma infection-related association with learning, pooled estimates suggested the presence of deficits for infected/non-dewormed vs. uninfected/praziquantel-treated children (n = 6 studies; SMD = -0.39, 95% CI: -0.70, -0.09). However, the magnitude of this association differed according to study design. Here, larger pooled standardized differences were realized for interventional studies (n = 2 studies; SMD = -0.79, 95% CI:-1.19, -0.39) than for observational studies (n = 4 studies; SMD = -0.18, 95% CI: -0.35, -0.01). Schistosoma infection was not significantly associated with performance on tests of reaction time (n = 6 studies; SMD = -0.06, 95% CI: -0.42, 0.30) or performance in tests of innate intelligence (n = 4 studies; SMD = -0.25, 95% CI: -0.57, 0.06). For all psychometrically-assessed cognitive domains, the overall findings were not sensitive to study quality. Of note, the majority of publications with psychometrically-evaluated endpoints had low risk of bias and produced pooled estimates similar in magnitude and direction to the overall results (Table 4).

Sensitivity analyses

High levels of heterogeneity were observed across included studies (P ≤ 0.03, I² ≥ 74.8%; Tables 2 and 4) for all outcome measures, but there was no evidence of undue influence by individual studies or of publication bias among the included studies (Egger’s test, all P-value ≥ 0.142, Table 2). Overall, inferences based on pooled estimates were generally insensitive to differences in study design (Table 4), to the exclusion of individual studies (S2 Table), or to the influence of publication year (S1 Fig).

Principal findings and interpretations

This systematic review and meta-analysis of the cognitive and educational impact of Schistosoma infection in school-aged children supports the hypothesis that infection is associated with reduced school-attendance, with deficits in scholastic achievement and deficits in memory and learning domains of psychometrically evaluated cognitive function. It has previously been conjectured that Schistosoma infection may affect school attendance, scholastic achievement, and cognitive function, either directly through via deposition of Schistosoma eggs within the central nervous system, via physical discomfort and subsequent distraction due to the presence of the worms, or indirectly, via iron-deficiency and malnutrition [3, 71, 72]. However, Schistosoma infection or non-treatment was not associated with performance in tests of innate intelligence or reaction time. Inferences based on most pooled estimates for psychometrically assessed endpoints were generally insensitive to study design, Schistosoma species, and risk of study bias. However, associations between infection and educational outcomes were sensitive to study design and study quality–especially estimates for impact on scholastic achievement. The association between infection and scholastic achievement was directionally similar and statistically robust regardless of Schistosoma species; however, average effects were substantially larger for S. haematobium compared to S. mansoni infection. Cohen’s criteria for effect size suggest that the average Schistosoma infection-related deficits in education, learning, and memory performance range from ‘small’ to ‘moderate’. S. haematobium infection was associated with relatively larger deficits in scholastic achievement.

Prior meta-analysis of four randomized controlled trials that evaluated cognitive impacts of STH infections–which sometimes co-occur with schistosomiasis, have reached a different conclusion about the cognitive and scholastic effects of STH infection and the impact of interval treatments for STH [25, 26]. Those reviews concluded that there was substantial evidence that deworming for soil-transmitted helminth infections does not yield a cognitive or educational benefit. We note that our approach differed from these prior STH-based meta-analyses on several grounds: a) we included both interventional and observational studies to take advantage of all research data available on this question, b) we evaluated Schistosoma-associated impacts on two domains of educational loss (attendance and achievement), c) we further defined domains of cognitive function based on psychometrically-assessed testing to include the following: learning, memory, attention, and intelligence and d) the intervention, as defined in this meta-analysis, denoted treatment for Schistosoma infection, whether or not the study was randomized.

We intentionally included data from all available epidemiologic studies–whether interventional or observational in design, in this first systematic review and meta-analysis of Schistosoma infection-related differences in cognitive and educational outcomes. The inclusion of all available evidence reflects current standards for clinical evidence-gathering to inform health policy, in order to shape clinical practice based on the ‘best available’ relevant information [23, 24]. Future randomized-controlled trials to address this question are expected to be limited in scope and may be considered ethically objectionable given the current widespread adoption of deworming for schistosomiasis and STH. The current adoption of ‘preventive chemotherapy’ guidelines has been based on helminth-associated adverse effects on anemia and child growth. In consequence, meta-analysis of available evidence, as performed in the present study, remains the most practical strategy to inform current policy. By this approach, we have identified ‘small to moderate’ infection-related deficits in education, learning, and memory performance using the Cohen’s criteria of effect size. However, such numerically small deficits (per Cohen’s criteria) may significantly underestimate clinical significance of infection for childhood development, as Schistosoma infection is an exposure affecting millions of children in endemic regions. Hence, small to moderate deficits at the individual level may amount to large and important differences in disease burden at the population level [73, 74]. Our interpretation of SMD estimates is ultimately grounded in the importance of summary measures for clarifying existing knowledge gaps regarding the relationship of Schistosoma infections to respective outcomes, and the expected benefit of systematically lowering infection-related deficits in the millions of children at risk.

As a limitation of our approach, we acknowledge the possibility of residual confounding and bias in the primary literature, given that majority of included studies were observational or non-randomized intervention trials. Only two of the 30 included studies used an RCT design, thus a sensitivity analysis based on RCT vs. non-RCT study design was not possible. However, as part of sensitivity analysis we evaluated the potential for differences in pooled estimates based on our expanded definition of intervention as including longitudinal studies that included praziquantel treatment. Investigation of pooled estimate sensitivity by region of study, the year of publication, and by risk of bias did not result in materially different statistical inferences. Specific investigation of publication bias suggests that any possible greater likelihood of publishing positive studies did not unduly influence observed findings. Our approach of including both interventional and non-interventional studies is consistent with theoretical and empirical evidence that meta-analyses based on observational studies generally produce estimates of effect similar to those from meta-analyses based on randomized controlled trials, and that a priori exclusion of observational studies in systematic reviews is inappropriate and inconsistent with the evidence-based medical decision-making approach [75, 76]. In addition, the often restrictive inclusion criteria and short follow-up duration in RCTs could easily result in outcomes largely different from when the same interventions are applied to a general population.

Despite the intuitive appeal of our cognitive domain based evaluation, we acknowledge the critique that our classification of psychometric instruments by domain required some level of subjectivity, especially for tools that capture performance across multiple domains. We explicitly identified the instruments used in each study and, based on literature description of the major cognitive domain assessed by each tool, combined related tools into four separate domains. Ultimately, each instrument was assigned to one cognitive domain only. We have described the logic for our choices in the supplementary information (S1 Table) to provide a basis for further discussion and give sufficient context for critical evaluation of our approach in developing future studies.

Our decision to combine psychometric evaluations of cognitive functions in four domains (based on the primary capacity being tested) is a strength of our empirical approach. By so doing, we recognize that Schistosoma infection may not have equal impact on all cognitive domains. For example, to the extent that innate intelligence is strongly influenced by fixed or heritable factors, we did not anticipate infection related differences on tests of ‘intelligence quotient’. Unlike intelligence tests, we considered the other tests of memory, learning, and reaction time to be more sensitive to infection, and thus modifiable by presence/non-treatment vs. absence/treatment of Schistosoma infection. Unexpectedly, reaction time was not associated with Schistosoma infections. However, our findings of infection-related reductions in learning and memory tests are consistent with our hypotheses that these cognitive domains are sensitive to adverse environmental perturbations–including Schistosoma infection.

Remaining gaps, and recommendations for future research

The finding of infection-related cognitive deficits and educational loss reported here is clinically and health policy-relevant for mitigating the cognitive and functional morbidities of Schistosoma spp. infection in children. The ‘small-to-moderate’ effects demonstrated may, in reality, be an underestimate of the lifetime impact of Schistosoma infection on personal performance (as affected via ultimately irreversible educational and cognitive losses). Typical epidemiologic studies necessarily include only a constrained portion of the relevant etiologic period. Among school-age children, Schistosoma infection is often effectively already chronic and/or recurrent, with reinfection rates extremely high in the absence of meaningful environmental interventions to reduce re-infection following treatment. The cumulative cognitive and educational impact of persistent infection may not be adequately captured by the relatively short-term investigations of treatment impact included in this meta-analysis.

In schistosomiasis-endemic regions, many children are infected by age two and remain chronically infected through school-age and late adolescence [9]. Under current national Schistosoma control treatment guidelines, preschool-age children are not treated as part of routine deworming programs for STH or schistosomiasis [1, 6, 30]. These children may therefore suffer cumulative damage to their health and function that is currently not reflected in most short-term study outcomes (or this meta-analysis). Of note, recent investigations have demonstrated the safety and efficacy of praziquantel for treatment of Schistosoma infection in preschool children [9]. The existence of an adverse developmental impact of Schistosoma infection on cognitive/educational domains would be a major justification for expanding the age-bracket of children who should be treated with praziquantel. Currently, evidence suggests that the timing of infection across a life path is especially consequential in terms of the severity of cognitive and physiologic impairments experienced [77, 78]. Future investigations evaluating the relative differences in cognitive outcomes for pre-school children with and without Schistosoma infection will be important for understanding the magnitude of potential impact of better prevention of Schistosoma infection.

It is currently unknown whether the cognitive and educational loss associated with Schistosoma infection can be reversed with treatment alone. The fact that infection often occurs in the context of malnutrition, coincident parasitic infections, and extreme poverty suggests that cognitive remediation efforts will need to be multi-faceted, using an integrated disease management framework. We expect that educational and cognitive interventions will be most effective if initiated earlier in life and that the package of interventions may need to include remedial instruction, the prevention of reinfection for treated/cured children, management of comorbid health conditions, and interventions for improvement of nutritional status.

Policy implications

Our investigation suggests that Schistosoma infection/non-treatment is associated with educational and cognitive loss. Our findings further suggest a definite cognitive and educational benefit of anti-schistosomal deworming among school-age children. Future complex intervention studies of early childhood interventions, focused on improving child well-being and cognitive potential, are needed to determine to what extent these observed deficits are preventable or reversible. Interventions that employ an integrated disease management framework will likely identify cost-efficiencies for leveraging existing disease and nutrition treatment programs in helminth-affected regions.