Longitudinal Studies in Pediatric Gastrointestinal Surgery: Theory and Practice

This thesis described the current state of longitudinal data analysis in the pediatric gastrointestinal (GI) surgical literature. Studies on two procedures were performed to gather the best available evidence for minimally invasive pediatric upper GI surgery. The studies were performed using the best possible design and analysis given the constraints.

Chapter one provided an introduction to the current state of methodology and statistics, especially regarding longitudinal data analysis, in the pediatric GI surgical literature. Several challenges in the reporting and analysis of the effects of pediatric surgical studies were summarized, such as the preponderance of case studies, the small number of patients in studies, ethical issues regarding a vulnerable population, poor reporting of methodology, and the suspicion that the statistical analysis of repeated measures is not always appropriate. In addition, the effects of two surgical procedures were described: anti-reflux surgery (LARS) for children with severe gastrointestinal reflux disease that does not respond to medication; and gastrostomy placement (GP) for children with severe feeding problems.

To identify potential methodological problems in the pediatric surgical literature, a literature review was performed on all scientific articles that reported on repeated measures in a pediatric GI surgical study in the years 2010-2019. The results were presented in chapter two. A large majority of the studies examined reported on the number of included patients, which was generally low (median sample size 40). Most also clearly stated the objective of the study and interpreted the results in that context. Poor adherence to reporting was found on several items related to data collection and analysis: sample size justification; reliability and validity of methods; statistical methods used; numbers of participants at each wave; and generalizability of results were not, or not clearly, reported in a majority of the articles examined. This information is crucial to the assessment of the quality of the study.

Furthermore, more than half the pediatric GI surgery studies classified as “longitudinal” or “cohort” in PubMed did not make efficient use of that design by collecting either time-to-event data or repeated measures on one or more outcomes. In the majority of the articles that did collect repeated measures, we observed inefficient, inappropriate or incorrect analysis of longitudinal data. The inefficient use of the cohort/longitudinal design identified in many publications on pediatric GI surgery has likely led to reduced power of the studies. In addition, inappropriate or incorrect analysis of repeated measures data at best make inefficient use of the available information, and at worst may have led to biased estimates of treatment effects. While these methodological issues are important in any medical study, they are especially so in studies on vulnerable (pediatric) populations….

The remainder of the summary can be found in the thesis.