Citrulline levels can not be used to delineate the severity of Chron's Disease for out-patients

Reference: Elkhatib, I. and A.L. Buchman, Plasma citrulline concentration as a marker for disease activity in patients with Crohn's disease. J Clin Gastroenterol, 2012.46(4): p. 308-10.

A study done on the use of citrulline levels to delineate the severity of Chron's disease.  

Brief Summary: Citrulline is an organic compound produced from the urea cycle, and it is thought to be an indicator for the severity of Chron’s Disease.  The main objective of this study was to determine if Citrulline could be used as an indicator for Chron’s.  The authors accrued 81 patients and obtained their Harvey-Bradshaw Index scores (HBI) to see if there was a significant correlation with their Citrulline levels.  The HBI scores are used to quantify the symptoms for a patient with Chron’s Disease.

Results: The author’s did not find a significant correlation between the HBI index and the Citrulline levels thus indicating that Citrulline cannot be used to delineate the severity of Chron’s Disease.  As the authors noted,  only outpatients were used and it’s possible that more severe Chron's cases with proximal bowel involvement (eg. jejunal or duodenal) might show a decrease in citrulline concentration.

Implications for Practice: Citrulline levels cannot be used to delineate the severity of Chron’s disease for out-patients.   

Discussion: It would be interesting to see what the citrulline levels would be for patients that have a more severe case of Chron’s.  It’s obviously hard to say, but my guess is that the nature of the statistical relationship between citruline and the HBI index would be significantly different for these patients.  Particularly, it would be interesting to look at the nature of the statistical relationship between citrulline levels and a whole slew of other clinical variables such as various bowel disruption variables.    

Commentary on Statistics and Study Design: The author’s used the simple Pearson's correlation measure to measure the correlation between the citrulline levels and the HBI index. However, the problem with using this simple correlation model is that other variables cannot be controlled for, and there were definitely other variables in this observational setting which could have been controlled and investigated on.
 

What should have been used is a multi-variate regression model.  By using this model, the outcome variable would have just simply been citrulline levels and the predictor variables would have been HBI index along with a bunch of other predictor variables such as post-operative recurrence, previous resection, etc. Remember, the advantage of using a multi-variate regression model is that it allows you to investigate the nature of the relationship between multiple predictor variables and a given outcome variable while controlling (keeping constant) all other predictor variables.  Hence, there was really no reason to get rid of the patients who had post-operative recurrence or previous resection.  These could have just been qualitative variables which were thrown into the multi-variate regression model and this would have taken care of any potential confounding effects.  One could then have also included a bunch of interaction variables to see if the nature of the relationship between HBI and the predictor variables depend on another variable.  As long as the patients are sampled independently, everything would have been OK. In fact, by throwing out the patients with post-operative recurrence abd previous resection , you are actually ‘losing’ valuable information.  You never want to throw out patients for the fear of a confounding variable amongst your patients.  Keep them in - you can't even assume a-priori that they would be confounding - and it would have been an interesting relationship to look at.