In this issue, Xanthakos et al., of Cincinnati, report a survey of pediatric gastroenterologists on their usual evaluation and management of healthy infants with rectal bleeding and contrast the results with their own experience with 22 similar infants (1). The authors' bias was that “allergic” colitis was empirically overdiagnosed by history and that empiric intervention with dietary restriction was unnecessary.
They followed 22 infants younger than 6 months of age with blood-streaked stools. All had normal growth, none were receiving hydrolyzed formula, all had negative stool cultures for pathogens, and all had normal anal examinations excluding fissures. Ten had diarrhea, eight had emesis or reflux, seven had nondiagnostic rashes, none were anemic, and 21 had a family history of atopy. No data were presented on RAST testing, stool white cells or serum albumin. All infants had flexible rectosigmoidoscopy to 15 cm without preparation or sedation. Pinch biopsies obtained at 5, 10 and 15 cm were reviewed blindly by a pathologist.
Of the 22 infants biopsied, 14 had “allergic colitis” using >6 eosinophils per high-power field as the diagnostic criterion. Seven were breastfed and seven were formula fed. Three infants received a histologic diagnosis of “non-specific colitis” and received no intervention. One of these progressed in the 9-week follow-up to infantile inflammatory bowel disease. Five infants had histologically normal biopsies despite clinical bleeding and visible colitis in two. On repeat biopsy, one of these infants developed findings of allergic colitis.
Formula-fed patients with allergic colitis were placed on extensively hydrolyzed formula. Six of the seven responded, whereas one responded only after starting L-amino acid formula. The seven breastfed infants continued nursing while the mother began a cow milk protein free diet. One of these infants was lost to follow-up and one continued to bleed until L-amino acid formula was started. Overall, 64% of the infants with bleeding were thought to have allergic colitis and thus received a dietary intervention.
A survey of 49 NASPGHAN members from Ohio asked what evaluation the responder felt was indicated in healthy infants with mild rectal bleeding. Forty-one physicians suggested empiric dietary intervention before invasive tests for allergic colitis. Only four responded that they would biopsy before initiating dietary alteration. The real quandary presented by these infants is this: should we document or should we treat empirically? The answer depends on how we view the failure rate of empiric therapy. Does a .640 batting average put empiric therapists at risk or does it place them in the Colitis Hall of Fame? Of the 22 infants in the study, only one had a diagnosis that eventually required intervention other than diet change. Ironically, that patient was not diagnosed on the initial biopsies. Webster's defines empiric therapy as a medical decision based on the theory that all knowledge originates with experience. If only we could modify the requirement for “all.”
The absence of a noninvasive diagnostic marker for this condition is a problem. A stained smear of the fecal mucus, interpreted by an experienced eye, may reveal fecal polymorphonuclear cells, suggesting inflammatory colitis rather than an anatomic bleeding source. Infants with proctocolitis often have borderline low serum albumin or protein as a result of protein losing enteropathy. This is especially prominent if there is lymphoid nodular hyperplasia of the rectum. As noted in this report, peripheral eosinophilia or total immunoglobulin E level was not a discriminatory diagnostic finding. Atopic features were not consistently found. Eczema was seen in only a small percentage of patients. The high prevalence of atopic family members in this study population was unusual.
The authors acknowledge the limitations of their diagnostic and therapeutic algorithm. Two parents refused enrollment because of concerns about the procedures; others declined follow-up. Even biopsies at three levels missed inflammatory features that can be quite focal. The role for multiple sequential biopsies has yet to be determined. Histologic inflammation can persist for weeks after the resolution of overt clinical symptoms.
The authors' concern over the expense of unnecessary hydrolyzed formula and L-amino acid formula is legitimate. Standard formulas cost about $6.00/1000 kcal while extensively hydrolyzed formula costs $11.00/1000 kcal and L-amino acid formula costs $16.00/1000 kcal. Six months of extensively hydrolyzed formula will cost a family approximately $1100.00 more than standard formula. Favoring empiric therapy, however, is the fact that some families are partially shielded by insurance coverage and that the cost of the sigmoidoscopy, pathologic interpretation, and facility fee exceeds $1500.00 at most facilities.
A pragmatic issue in care of these infants is expeditious access. I asked one of my office staff to impersonate a concerned mother and call five referral centers within 2 hours of her home requesting evaluation for her 3-month-old healthy baby with blood streaks in the stool. All calls were handled by a receptionist who offered either an appointment in 4 to 10 weeks or an evaluation in an emergency room. Should we consider such delay an advantage? Six of the 22 infants in Cincinnati would have stopped bleeding before they got an appointment. This was obviously a created situation. In fact, most referrals come from primary care providers who refer only after their empiric therapy has failed.
What about the breastfed infant with benign bleeding? In all the studies referenced by the authors, breast fed infants accounted for 50% to 70% of the patients. In my review of 95 exclusively breast fed infants published in the Journal of Pediatric Gastroenterology and Nutrition in January, 2000 (2), elimination of a single dietary protein in mother's diet led to clinical resolution in 79 patients. Five mothers had to eliminate two or more proteins. In 11 infants, bleeding ceased within 72 hours of starting protein hydrolysate formula and resumed with resumption of nursing despite multiple maternal dietary interventions. Of the 95 nursing mothers, only two stopped nursing sooner than they had planned. In 62 of 95, cow milk in the maternal diet was the offending antigen. All infants tolerated the offending protein in their diet by the age of 9 months.
What is the risk of continued nursing in the infant with eosinophilic proctocolitis who continues bleeding? We found that 11 of 95 mothers could not identify the offending protein and an additional 10 chose not to maintain their restricted diet even when the offending protein was known. Twenty of these 21 mothers continued to nurse for longer than 4 months despite the infants' persistent intermittent bleeding. Six of the infants became mildly anemic despite iron supplementation; none became malnourished.
What is the long-term prognosis of these infants? David Hill reported no long-term intestinal problems in 13 infants followed for 10 years (3). I have now followed 74 breastfed infants with proctocolitis for more than 10 years and have found no documented intestinal disease (unpublished data). Eight infants born subsequently to these mothers have developed eosinophilic proctocolitis. Six of the eight appear sensitive to the same protein. There are no published studies on the long-term outcome of formula-fed infants with eosinophilic proctocolitis who remained on the offending protein. My personal experience is that colitis seems to progress in severity when intervention is delayed. I have observed that inadvertent protein exposure can provoke bleeding within 48 hours, especially when soy protein is the offending antigen.
I close with a pet peeve about the classification of dietary protein-induced enteric disease. By present classification, adopted in 1998 (4), none of the infants in this study had “allergic” colitis. There are no data to prove that protein-initiated, immunoglobulin E-mediated pathophysiology is at work here. The eosinophil is the marker of the inflammatory response rather than the confirmation of allergy. Thus I prefer the term eosinophilic proctocolitis for infants such as those in this study. The conscience of Cincinnati has spoken and we should listen. Even ye olde experience-laden empiricist agrees that prospective multicenter studies addressing the natural history of this common condition are required.
1. Xanthakos S, et al. Allergic colitis and rectal bleeding. J Pediatr Gastroenterol Nutr
2. Lake AM. Food-induced eosinophilic proctocolitis. J Pediatr Gastroenterol Nutr
3. Hill DJ, Ford RP, Shelton, MJ. A study of 100 infants with cow's milk allergy. Clin Rev Allergy
4. Sampson HA, Anderson JA. Summary and recommendations: Classification of gastrointestinal manifestations due to immunologic reactions to foods in infants and young children. J Pediatr Gastroenterol Nutr