Reintam and colleagues recently proposed for the first time a score of gastrointestinal failure (GIF) 1, and confirmed its association with poor outcome. The authors should be congratulated for their attempt, and for the combination with the already existing well-validated Sequential Organ Failure Assessment (SOFA) score 2.

Organ dysfunction is an evolving state: its expression and conditions vary. An ideal score should be specific, be sensitive, be objective, reflect a comprehensive measure of the system, be inexpensive, be abnormal in one direction, and be a continuous variable 3. The proposed items do not fulfill these criteria. In addition, the items' applicability in different diagnostic categories is questionable. Furthermore, the study population, which is small for the elaboration of a score, is not representative of a general intensive care unit population, with very few cases of respiratory failure or shock, an unusual high proportion of comas (30%), and high mortality.

Food intolerance assessment depends on unit protocols. Failure of progression should therefore be assessed as per protocol rather than as 50% of needs. Withholding feeds does not reflect the patient's physiological alterations, but is a medical decision not necessarily reflecting disease severity.

A major problem is to consider the GIF score as a continuous variable, which it is not. This oversight explains the limited impact on the receiver operating characteristic curve of the score alone or in combination. A GIF score ≥ 3 was associated with a sharp increase of intensive care unit mortality: this reflects the strong impact of intra-abdominal hypertension (IAH) and abdominal compartment syndrome on mortality 4. Indeed, the fact that food intolerance alone is not an independent predictor of outcome supports this interpretation, questioning the rationale for its aggressive prevention or treatment. The GIF score should have been used as a categorical variable, with a dichotomization (≥ 3 or <3 points), enabling answering the question of whether the score itself had prognostic value or whether only IAH and abdominal compartment syndrome were independent predictors. Alternatively, testing separately IAH, abdominal compartment syndrome, and GIF as dichotomized variables in the regression model, and calculation of the odds ratio adjusting for other covariates, would have shown whether adding the GIF score to the SOFA score was contributive to prognosis. Finally please note that there is an unlucky typographic error in Table 2: several numbers ± standard deviation should be numbers and percentages 5.

We need to better characterize gut function. The score should include a combination of variables, as the gut is involved to a variable extent in different diagnostic categories. Table 1 proposes symptoms and signs that can apply across medical and surgical conditions, and which we already use for gut assessment in our department (unpublished data). The inclusion of IAH is essential in relevant conditions (for example, major trauma/burns, major vascular surgery, transplantation) 4. Elevated lactatemia raises the suspicion of intestinal ischemia; however, elevation of lactate is nonspecific and late. Gastric residues remain, despite methodological problems, the gold standard of feeding tolerance assessment 6: an increasing volume is clinically relevant. Among motility, both constipation and diarrhea – two forms of gut failure – should be included. Further prospective multicentric investigation of the validity of the combination of continuous items is required to finally characterize this invisible organ.

Authors' response

Abbreviations

Competing interests