Inappropriate Methods May Lead to Misleading Conclusions From the Meta-analysis of Observational Studies.

Abstract

To the Editors: We read the article by Renko and colleagues with the outmost interest.1 This article discusses an important and relevant aspect of urinary tract infection (UTI) for pediatricians. The authors in this article systematically reviewed and synthesized data on risk factors for UTI in children. Twenty-four studies were included in the meta-analysis. Authors found that circumcision and breastfeeding reduce the risk of UTI, and poor fluid intake, infrequent voiding, and being overweight or obese increase the risk of UTI recurrence in children. More importantly, the authors also observed that obstipation and high-grade vesicoureteric reflux do not increase the risk of recurrent UTIs. Overall, this article addresses an important issue for clinicians, but we came across a few methodological concerns, which may affect the validity of the findings of this meta-analysis. It is usually recommended that the literature search be comprehensive (at least 2–3 bibliographic electronic databases) before conducting a meta-analysis. Systematic reviews of observational studies require an even more robust literature search.2,3 Authors performed a literature search only in 1 database (PubMed); hence, the likely possibility that few studies may have been missed putting this systematic review at higher risk of publication bias. In this article, authors did not report publication bias, which was not discussed under limitations. A few missed studies (Gondim et al. 2018, Shaikh et al. 2016, Conway et al. 2007, and Cho et al. 2005) that assess risk factors for UTI are missed. High-grade VUR and bladder-bowel dysfunction (BBD) are 2 widely recognized and reported risk factors for UTI. Still, the present meta-analysis failed to demonstrate these risk factors for recurrence of UTI, as the literature search was not extensive and missed studies. The authors state that they have followed the checklist of the Meta-analysis of Observational Studies in Epidemiology group (MOOSE) but failed to report the search strategy for literature search, which is an essential component per this MOOSE guidance. In the current risk-stratification model for recurrence of UTI, age, sex, VUR severity, BBD, and circumcision status are considered important factors.4 However, the authors in this meta-analysis did not assess age and sex as risk factors and failed to show the other 2 factors (BBD and VUR) as risk predictors for recurrent UTI in children. Similarly, some studies (Ellison et al. 2018, Dubrovsky et al. 2015, Braga et al. 2015, Alsaywid et al. 2010, Nayir et al. 2001) assessing the circumcision status as a risk factor for UTI are also missed.5 These findings are likely caused by a suboptimal literature search and an inappropriate study selection process. The authors mention low heterogeneity in the text, but it has not been reported for most risk factors except I2 –7.5% for circumcision status. Furthermore, the authors do not explicitly report what was considered severe congenital renal abnormalities as they used it as one of the exclusion criteria. Although they included studies with VUR grade IV and V, we feel that these limitations could have been better discussed in the limitation section because these may have had a significant impact on this meta-analysis’s finding.