We will conduct a systematic review and meta-analysis on the association of sex/gender-with mortality in the pediatric sepsis population. This review protocol will follow the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols guidelines (18) and was submitted for registration in the International Prospective Register of Systematic Reviews (PROSPERO).(19) As this is a systematic review of the literature, no ethics approval will be required.
We are aware that even though most studies report on sex or gender differences, they actually report on biological or reported sex as opposed to gender as defined above. We will therefore summarize our findings under the term “sex” which will reflect sex differences and will not specifically evaluate for gender differences. We will report on the number of studies using each term (sex and gender) and whether or not they justified the use of the reported term.
Eligibility Criteria
We will include prospective and retrospective cohort studies, case-control studies and randomized trials published in English during a 17-year period (1st January 2005 – 31st December 2021). We have limited our review to English language studies based on previous sepsis reviews showing that this strategy resulted in omission of only 0.4% of studies.(20)
We chose this time frame due to the publication of the International Pediatric Sepsis Definition Consensus Conference statement in 2005.(21) This will allow sepsis definitions to be more consistent for data collection and is aligned with other sepsis reviews.(22) We will exclude systematic reviews, abstracts, case reports and narrative reviews from our study. We will only include studies which provide mortality outcomes according to sex. We will screen references of included studies to ensure no relevant publications are missed for analysis.
The study question was defined based on CHARMS (checklist for critical appraisal and data extraction for systematic reviews of prediction modelling studies).(23)
Population
Our study population will include children > 37 weeks post gestational age to < 18 years of age diagnosed with sepsis, severe sepsis or septic shock (see sepsis definitions(22)) admitted to a pediatric intensive care unit (PICU), intensive care unit (ICU) or have an explicit diagnosis of organ dysfunction. In order to conduct a pragmatic and inclusive systematic review, studies with the terms sepsis, severe sepsis, and septic shock will be included independent of the classification system used (IPSCC 2005, Sepsis-3, WHO definition, Sepsis-1 and Sepsis-2 definitions). If patients are diagnosed with sepsis and hospitalized in the ICU, we assume that infection was severe enough to meet the previously mentioned sepsis criteria(21). However, we will also include studies that were conducted in a setting without a PICU, if the criteria for sepsis with organ dysfunction were met.
If study data allows, we will divide patients into age groups (infant, pre-school, school age, adolescent), which will be evaluated for sex-differences per age group. We will exclude premature infants and studies conducted in a neonatal ICU (NICU) from our analysis, as admission to the ICU in this group may be more reflective of gestational age than severity of infection. In addition, preterm infants and newborns who have never left the hospital represent patient groups of different vulnerability, epidemiology, and outcomes, compared to the other pediatric age groups. Studies of all sepsis pathogens (viruses, fungi, parasites, bacteria etc.) and of children with inconclusive microbiological findings will be included. We will include only studies which provide analyses or reporting according to sex. Studies on exclusively adult patients or adult studies in which the pediatric data is not separately reported will be excluded.
Index Comparator
The main comparison will be between male and female sex.
Outcomes
The primary outcome will be hospital mortality. Secondary outcomes will be PICU and hospital length of stay (LOS).
Data Sources
We will identify eligible studies by searching the following databases: MEDLINE and Embase.
We will perform a hand search of references of the primary studies and systematic reviews for relevant studies to be included in our analysis.
Search Strategy
We have developed a search strategy with the help of an information specialist from the University Library Zurich. The following search terms were used: “sepsis, severe sepsis, septic shock, child, pediatric, mortality”. The detailed search strategy is outlined in the appendix.
Study Selection and Screening Process
The titles found in the primary search will be screened with the help of a web-based systematic review platform (Covidence). Citations will be screened for eligibility based on title, abstract and full text. Screening of citations at each level will be performed by two independent reviewers. Conflicts will be resolved by a third reviewer. Titles rejected by both reviewers will be excluded and the remaining titles will undergo full text screening. Each full-text article will be screened by two investigators for inclusion in the final dataset, which will then undergo data extraction. Conflicts will be resolved by a third reviewer. Kappa scores on the level of agreement reached between the reviewers at each stage of screening will be calculated and reported. We will document the rationale for inclusion and exclusion of studies.
Data Extraction and Management
Data from included full-text articles will be extracted by two reviewers and inserted into an electronic case report form (eCRF) in Redcap. We will collect data on the following potential covariates if available in the included studies: congenital disease or any other non-congenital major comorbidity (such as oncological diagnosis), race/ethnicity, type of infection, severity of illness scores (Pediatric Index of Mortality II(24), PRISM(25) score) at admission. If sufficient information for data extraction is not evident in the full text, we will contact authors for further information on study data.
Data Items
Data extraction will include information related to study characteristics (title, authors, year of publication, country, journal, study design, sample size, and inclusion/exclusion criteria); population characteristics (age, sex, ethnicity, admission diagnosis, location of the admission within the hospital), exposure (sepsis, severe sepsis, septic shock), covariates (comorbidities: chronic neurological, respiratory or cardiac disorders, oncology and transplant patients, type of infection, severity of illness scores (Pediatric Index of Mortality II(24), PRISM(25) at admission); and outcomes ( hospital mortality, PICU and hospital LOS). If needed, the authors of the original articles will be contacted to obtain missing data.
Assessment of quality and risk of bias in included studies
Since this analysis is based on existing research, publication bias is possible when performing a systematic review. Selection bias will be addressed with a systematic and reproducible search by an information specialist. References in included studies will be screened to further reduce the risk of selection bias.
All included studies will be evaluated with respect to their risk of different biases (e.g., selection, attrition, detection bias) according to study design. We will assess the quality in selected studies using the Quality in Prognosis Studies (QUIPS) tool for assessment of risk of bias.(26) Risk of bias assessment will be reported as “low risk”, “intermediate risk” or “high risk”. Risk of bias will be assessed by two independent raters, and discrepancies will be resolved by involving a third party.
A subgroup analysis will be performed to assess for a subgroup effect. Should this be the case, we will perform a separate sensitivity analysis for high ROB Studies.
Data Synthesis and Analysis
Statistical methods will include study-level descriptive statistics the studies included in the systematic review. It is anticipated that the majority of included studies will not be focused on sex as a prognostic factor for mortality in sepsis but will instead simply report sex proportions in those who did and did not survive. As such, our analysis will simply be to determine if there is an association between sex and mortality in pediatric sepsis and will serve as a basis for future prognostic models. Forest plots will be used to visualize the primary and secondary outcomes across studies. If the same secondary outcome is reported in two or more studies or study arms, a meta-analysis will be calculated with a random effects model as we anticipate substantial heterogeneity between studies. If possible, meta- regression models will be fitted to quantify the association between covariates and the primary and secondary outcomes. Funnel plots will be shown to address reporting/publication bias if at least ten studies report on the primary outcome. Depending on the outcome distribution, different effect measures will be meta-analyzed: proportions for in-hospital mortality (and PICU), and mean or median durations for hospital LOS. Statistical heterogeneity will be reported by an I2-statistic and Cochrane’s Q.
We will provide a narrative synthesis of our results if statistical analysis is not possible due to a low number of studies or heterogeneous outcomes. If four or more studies report a given covariate, we will perform subgroup analysis by age, congenital disease or any other non-congenital major comorbidity (such as oncological diagnosis), ethnicity, type of infection, severity of illness scores (Pediatric Index of Mortality II(24), PRISM(25) score) at admission. A sensitivity analysis will be conducted for the primary outcome, including all studies that have been rated as low risk of bias.