Gastrointestinal bleeding in patients with Henoch–Schoenlein purpura

Abstract

We read with much pleasure the recent research article by Hong et al. entitled “Neutrophil-to-lymphocyte ratio to predict gastrointestinal bleeding in Henoch–Schoenlein purpura (HSP)”. Children with HSP are notoriously at risk of sometimes severe abdominal complications, such as intussusception (by far the most frequent complication), bowel ischemia, intestinal perforation or massive lower or upper gastrointestinal haemorrhage. In a retrospective study, the authors identified neutrophil-to-lymphocyte ratio as a potential predictive biomarker of gastrointestinal bleeding in children with HSP. Given the aforementioned potential digestive complications of HSP, the availability of a simple laboratory test to predict severe gastrointestinal bleeding is more than welcome. Nonetheless, before using such a biomarker on a larger scale, we have a few minor concerns with some methodological aspects of this paper, especially the identification and selection of the patients, the definition of the outcome, and the prediction model definition. This study is based on an archival research design. The process of case identification, however, is poorly described. It would be interesting to know how cases were identified: that is, from personal files or from the central medical archives of the hospital, via the International Classification of Diseases (ICD) codes for example. This information would help in the assessment of the risk of selection bias or recall bias. Also, this would allow identification of any missing charts (again with potential biases). In addition, the sampling strategy is not clearly described: was it a convenience, random or systematic sampling. Although not obligatory, a sample size for correlation could have been calculated as well. The description of case selection is excellent for the diagnosis of HSP. In contrast, patient outcome in terms of gastrointestinal involvement is less clear. In this original research, the definition of gastrointestinal bleeding ranges from massive upper or lower digestive bleeding to the isolated finding of positive stool occult blood test. It is therefore impossible from the description “gastrointestinal (GI) involvement” or “GI bleeding” reported by the authors to ascertain how many children had massive gastrointestinal bleeding and how many had positive stool occult blood test. A clinical distinction between these abdominal complications is important because their consequences are dramatically different, ranging from observation to transfusion, intensive care admission, or surgical exploration. Finally, we have a minor remark regarding the prediction model. The study design, with a simultaneous measurement of an independent (neutrophil-to-lymphocyte ratio) and a dependent variable (gastrointestinal bleeding) is related to a diagnostic prediction model, and not a prognostic model. Based on these findings, physicians should not use neutrophil-to-lymphocyte ratio to predict the occurrence of abdominal bleeding in the future, which is rightfully acknowledged by the authors. Further studies might help clarify this last aspect and validate these interesting preliminary findings.