Implementation procedure
To perform this meta-analysis, the diseases of interest are neutropic parasitic diseases: malaria, cysticercosis, toxoplasmosis, HAT, Chagas disease, and human toxocariasis [9, 13]. These are forms of parasitoses that have a predilection for infesting the central nervous system and which can result in neurological disorders. Mental disorders of interest included: anxiety, depression, bipolar disorder, and schizophrenia [10, 11]. Mental disorders are outlined as medical conditions that interfere with thinking, feeling, mood, communication, and daily functioning, which typically lead to a reduced ability to address routine daily activities, such as working or raising a family [36].
Since this meta-analysis is the continuation of a previously published study on the association of mental disorders and chronic physical diseases in developing and emerging countries, its shares the same methods and research strategy [37]. Its protocol was recorded in PROSPERO, accessible via the following link: http://www.crd.york.ac.uk/PROSPERO and registered under CRD42017056521. It follows the recommended methodology for the meta-analysis of observational studies [38] and was performed in accordance with Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [39].
Research strategy
The search for articles was conducted through four databases: Medline, Embase, Lilacs, and IENT (database of the Institute of Epidemiology and Tropical Neurology of the University of Limoges in France: http://www-ient.unilim.fr/). LOD, the principal investigator, conducted article searches on these databases from February to May 2017 without linguistic or date restrictions, using the same research equation built on Medline [37].
(“Depressive Disorder”[Mesh] OR “Depression” [Mesh] OR “Anxiety Disorders”[Mesh] OR “Anxiety” [Mesh] OR “Bipolar Disorder”[Mesh] OR “Schizophrenia”[Mesh]) AND (“Diabetes metillus”[Mesh] OR “Obesity”[Mesh] OR “Neoplasms”[Mesh] OR “Cardiovascular Diseases”[Mesh] OR “Pulmonary Disease, Chronic Obstructive”[Mesh] OR “Malaria”[Mesh] OR “Cysticercosis”[Mesh] OR “Toxoplasmosis”[Mesh] OR “Toxocariasis”[Mesh] OR “Trypanosomiasis”[Mesh] OR “Chagas Disease”[Mesh]) AND (“Name of a country”[Mesh]).
This equation, initially built for both studies, allowed us to search articles on: associations of mental disorders and chronic physical diseases, which have been published in another meta-analysis; and associations of mental disorders and neurotropic parasitic diseases, that we discuss in this article.
In Lilacs and Embase, the keywords employed to build the research equation were identical. In total, 139 different equations corresponding to the 139 developing and emerging countries studied were used to search for articles on three databases (Medline, Embase, and Lilacs). When searching on the IENT database, articles were searched using the additional terms “comorbidity” or “comorbidity in mental health.” This methodology was adopted for the latter database because it is specifically dedicated to research on neurotropic parasitoses in developing and emerging countries. Finally, the registration and selection of articles throughout this work was done through the Zotéro bibliographical management software package.
Inclusion criteria and selection of articles
Every article included in this meta-analysis had to be: an article with full text available; a cross-sectional or analytical study; conducted among adults, male and female, of all ages (age ≥ fifteen years); and a study involving either only hospitalised subjects or only non-hospitalised subjects (but not hospitalised subjects and non-hospitalised subjects at the same time). Every article also had to specify the method used to diagnose the diseases. In this study, by non-hospitalised, we mean patients who reside in the community or who attended a health centre (clinic or hospital) for care but did not remain at the health centre for one or more nights. On the other hand, hospitalised patients are those who have stayed in a health care centre for one or more nights. The cross-sectional studies included in this meta-analysis had to each present the prevalence, or the data from which it might be calculated. Analytical studies had to offer the association measures, or the data from which they might be calculated.
Data extraction and assessment of article quality
The data were extracted from each article by LOD (the principal investigator). This included: reference, title, country and continent, the study type and study population, the original disease, the associated disease searched for and its method of diagnosis, the prevalence or measures of association of the disease searched for (or the data needed to calculate them), as well as the sample size, sex ratio, and mean subjects age.
The assessment of the studies quality was performed independently by two investigators (LOD and PEB) using the revised Downs and Black assessment grid. Every article was assigned a score by each investigator [40, 41] and, therefore, the final score was set by mutual agreement after examination of the ratings. In the event of a disagreement, the expert opinion of a third investigator (PMP) was requested, without this investigator knowing the scores given by the first two investigators. The final score was given by mutual agreement between the three investigators.
Statistical analyses
The software system Comprehensive Meta-Analysis (CMA) Version 3.0 [42] was used for the analyses. The heterogeneity of the selected studies was assessed using Q and I2 statistical tests [43, 44]. The value of I2 was considered such that I² = 0 showed a lack of heterogeneity, I² < 0.25 showed low heterogeneity, I² between 0.25 and 0.5 showed moderate heterogeneity, and I² > 0.5 showed significant heterogeneity. Pooled estimates were then calculated using the DerSimonian-Laird random-effects technique [45], with the results obtained displayed on a forest plot [46] with a significance threshold of 5%. Afterwards, the investigation for publication bias was performed by constructing a funnel plot, a Duval and Tweedie trim and fill test check [45], and an Egger’s regression [47]. To assess the robustness of the principal results of our meta-analysis, we carried out a sensitivity analysis by removing the study with the greatest weighting as well as the studies of lower quality within the pooled studies. The study variables “original disease,” “associated disease,” “subjects type,” and “continent” were used to perform subgroup analyses.