Background Idiopathic iron-deficiency anemia in adults is assumed to be the result of occult chronic blood loss from the gastrointestinal tract. The aim of this study was to determine an effective clinical strategy for managing this common clinical problem.
Methods We prospectively studied 100 consecutive patients with iron-deficiency anemia, using colonoscopy and esophagogastroduodenoscopy and, in patients with negative endoscopic studies, enteroclysis (radiographic examination of the small intestine).
Results Gastrointestinal endoscopy revealed at least one lesion potentially responsible for blood loss in 62 of the 100 patients. Endoscopic examination of the upper gastrointestinal tract showed a bleeding source in 36 patients, and colonoscopy showed a lesion in 25; 1 patient had lesions in both the upper and lower gastrointestinal tracts. The most common abnormality in the upper gastrointestinal tract was peptic ulceration (duodenal ulcer in 11 patients, gastric ulcer in 5, and anastomotic ulcer in 3). Cancers, detected in 11 patients, were the most common colonic lesions. Enteroclysis was performed in 26 of the 38 patients with negative endoscopic studies, and the results were normal in all instances. Symptoms at a specific site in the gastrointestinal tract were predictive of disease in the corresponding portion of the bowel. In addition, the combination of positive tests for fecal occult blood and symptoms in the lower gastrointestinal tract had a positive predictive value of 86 percent for detecting a lesion in the lower gastrointestinal tract.
Conclusions Gastrointestinal lesions (in both the upper gastrointestinal tract and the colon) are frequently found in patients with iron-deficiency anemia. Since site-specific symptoms are predictive of abnormalities in the corresponding portion of the bowel, the initial evaluation should be directed by the location of the symptoms. Concomitant lesions of the upper and lower gastrointestinal tract are rare; thus, detection of a likely source of blood loss during the initial examination may obviate the need for further procedures.
Methods
All patients with suspected iron-deficiency anemia who were referred to the gastroenterology service from June 1990 through November 1992 were prospectively screened for enrollment in the study. Iron-deficiency anemia was defined as a hemoglobin concentration 
12.5 g per deciliter (7.8 mmol per liter) for men (normal range, 13.5 to 17.5 [8.4 to 10.9]) and 10.6 g per deciliter (6.6 mmol per liter) for women (normal range, 11.6 to 15.8 [7.2 to 9.8]), accompanied by at least one of the following laboratory values consistent with iron deficiency: a serum iron concentration 45 µg per deciliter (8.1 µmol per liter; normal range, 50 to 150 [9.0 to 27.0]) with a transferrin saturation no higher than 10 percent (normal range, 16 to 60 percent), a serum ferritin concentration 20 µg per liter for men (normal range, 20 to 450) and 10 µg per liter for women (normal range, 10 to 250), or the absence of iron stores in bone marrow-biopsy specimens. The criteria for excluding patients from the study were an obvious cause of blood loss (e.g., epistaxis or heavy menstrual flow), active gastrointestinal hemorrhage, severe cardiopulmonary disease, a serum transferrin concentration 2.16 g per liter (normal range, 2.16 to 3.99), reported or suspected pica, age under 18 years, and an inability to give informed consent for endoscopy.
During the study period, 163 patients were referred to the gastroenterology service for evaluation of purported iron-deficiency anemia. Thirty patients were ineligible for the study because they did not meet the criteria for iron-deficiency anemia. Another 33 patients were ineligible for the following reasons. Iron studies were not completed in 11 patients who had been referred for suspected iron-deficiency anemia on the basis of red-cell indexes: 4 patients began iron therapy before the blood tests were performed, 3 had tests only for serum iron, 2 were lost to follow-up, and 2 declined further blood studies. Sixteen patients were excluded because their serum transferrin values were less than or equal to 2.16 g per liter, suggesting that they had anemia associated with chronic disease. Six patients with documented iron-deficiency anemia were excluded because they were unable to give informed consent (two patients), had an obvious nongastrointestinal source of blood loss (two patients), or had unstable cardiovascular status (two patients).
Detailed clinical data were obtained for all patients enrolled in the study, with an emphasis on gastrointestinal symptoms. The specific symptoms of the upper gastrointestinal tract that patients were asked about included dysphagia, odynophagia, heartburn, dyspepsia, nausea, vomiting, anorexia, and upper abdominal pain related to eating or relieved by antacids. The specific lower gastrointestinal symptoms inquired about included hematochezia, a change in bowel habits, diarrhea, constipation, and lower abdominal pain that was colicky or changed with the passage of stool. Patients were also asked about the use of aspirin, other nonsteroidal antiinflammatory drugs, and alcohol. Before the endoscopic evaluation was performed, stool samples obtained by the patient or by digital rectal examination were tested for occult blood with the use of Hemoccult II slides (SmithKline, Sunnyvale, Calif.).
Patients were sedated for sequential endoscopic procedures with intravenous midazolam, meperidine, or both, while vital signs and oxygen saturation were monitored and electrocardiography was performed. The colonoscopy was performed first, followed immediately by esophagogastroduodenoscopy with the same anesthetic. The cecum was reached in all patients. There were no complications of the endoscopic procedures. If both upper and lower endoscopic examinations were negative, patients were referred to the radiology department for enteroclysis, which was performed by fluoroscopic intubation of the duodenum and instillation of barium and methylcellulose into the small intestine11.
For the colonoscopic examination, the following lesions were judged to be sources of substantial occult blood loss: a carcinoma, one or more adenomatous polyps over 1.5 cm in diameter, vascular ectasias that numbered five or more or were at least 8 mm in diameter, active colitis, or a colonic ulcer more than 1 cm in diameter. Diverticula were judged not to be a source of chronic blood loss. For the endoscopic examination of the upper gastrointestinal tract, the lesions considered to be sources of blood loss included esophagitis with erosions involving at least 5 percent of the mucosal surface of the esophagus,12 erosive gastritis (defined as at least 50 erosions of 1 mm or more with white bases encircled by erythema),13 a duodenal or gastric ulcer more than 1 cm in diameter, a carcinoma, adenomatous polyps, and vascular ectasias (the last two as described above). Hiatal hernias and esophageal varices were not considered to be causes of chronic blood loss. In the patients undergoing enteroclysis, mass lesions more than 1 cm in diameter, ulcerations, and extensive mucosal inflammatory changes were judged to be probable sources of blood loss. Endoscopic evidence of bleeding was noted, including hemorrhagic mucosa and a visible vessel in an ulcer.
Statistical analyses were performed with SAS software14,15. The upper confidence limit for a binomial parameter was computed exactly16.
Results
Clinical Features
The study group consisted of 73 outpatients and 27 inpatients, 51 of whom were women. The mean (±SD) age was 60 ±14 years, with a range of 20 to 85 years. Nine women were menstruating; none had a history of unusual or heavy flow. All the patients were anemic, with serum ferritin levels or transferrin-saturation values that were consistent with iron deficiency (Table 1). Bone marrow-biopsy specimens from two patients had no stainable iron.
|
Endoscopic Findings
A substantial gastrointestinal lesion was detected in 62 of the 100 patients (Table 2). Colonoscopy revealed a likely bleeding site in 25 patients, whereas esophagogastroduodenoscopy demonstrated a lesion in 36 patients; 1 patient had both a gastric and a colonic lesion. Cancer was the most common lesion in the colon, accounting for 11 (8 in the right colon and 3 in the left) of the 26 colonic lesions (42 percent). Three patients had Dukes class B cancers, three had Dukes class D, and five had Dukes class C17. All the colon cancers were ulcerated. Five patients with cancer had symptoms. Among the 10 patients with colon cancer who had valid stool guaiac tests (see below), 8 had positive tests. Only one patient with colon cancer both was asymptomatic and had a negative test for occult blood. Of the five patients with vascular ectasias (three cecal and two left colonic), four had multiple lesions; all five patients were treated endoscopically. Of the two patients with colitis, one had radiation colitis, and the other idiopathic ulcerative colitis. One of the cecal ulcers was idiopathic, and one was tuberculous. Fifteen patients with colonic lesions had stigmata of recent bleeding (11 of the 15 had cancer, and 1 each had vascular ectasia, tuberculous cecal ulceration, radiation colitis, and a polyp).
|
Thirteen patients had unexplained weight loss: seven with upper gastrointestinal abnormalities, four with cancer of the right colon, and two with no lesions. The mean weight loss was 13 kg (range, 2 to 36), 12 kg (range, 2 to 18), and 7 kg (range, 2 to 12) in those with upper gastrointestinal lesions, colon cancer, and no lesions, respectively.
Twenty-one patients reported long-term use of substances that irritate the gastrointestinal mucosa, including nonsteroidal antiinflammatory drugs other than aspirin (11 patients), aspirin (4 patients, none of whom took more than 325 mg per day), and alcohol (6 patients). Seven of the patients taking nonsteroidal antiinflammatory drugs had no identifiable lesions, two had lesions in the upper gastrointestinal tract (gastritis in both cases), and two had lesions in the lower gastrointestinal tract. One patient taking aspirin had colon cancer, and three had no detectable lesions. Of the six patients who drank alcohol regularly, three had lesions in the upper gastrointestinal tract (esophagitis, vascular ectasia, and duodenal ulcer in one each), and three had no lesions.
Radiographic Findings
Thirty-eight patients with negative endoscopic examinations of the upper and lower gastrointestinal tract were referred to the radiology service for enteroclysis. Twenty-six underwent this examination, and three additional patients underwent a small-bowel follow-through study after swallowing barium (because of difficulty with insertion of the enteroclysis perfusion tube). No lesions were identified in any of these patients.
Statistical Analysis
Values for hemoglobin, mean corpuscular volume, serum ferritin, and transferrin saturation did not differ significantly between patients with positive endoscopic studies and those with negative endoscopic studies.
Site-specific symptoms were significantly associated with lesions in the portion of the gastrointestinal tract where the symptoms originated, and these symptoms were predictive of abnormal findings (Table 3 and Table 4). Although site-specific symptoms were strongly associated with site-specific lesions, the absence of symptoms did not exclude the possibility of detecting a lesion. For the statistical analysis, a valid test for fecal occult blood was defined as at least one positive test or at least three negative tests; 90 patients had valid tests for stool blood. Stool testing alone had a relatively low sensitivity for detecting disease and a positive predictive value of 75 percent overall. Among patients with site-specific symptoms and positive tests for fecal occult blood, the positive predictive value for detecting lesions in the lower and upper gastrointestinal tracts increased to 86 percent and 82 percent, respectively.
|
|
Patients with Negative Endoscopic and Radiographic Studies
Lesions considered unlikely to be responsible for chronic blood loss in the 38 patients with negative endoscopic and radiographic studies included small polyps (1 cm in 1 patient, 0.5 cm in 1 patient, and less than 4 mm in 4 patients), hemorrhoids alone in 4 patients, antral hyperemia in 3 patients, small esophageal varices in 2 patients, and minimal esophagitis in 1 patient. No stigmata of bleeding were identified in connection with any of these lesions.
All 38 patients were empirically treated with oral iron; 36 of them were seen for follow-up tests 6 to 38 months later (mean, 20 months). Iron-deficiency anemia had resolved (defined as an increase in hemoglobin above the lower limit of normal) in 30 of the 36 patients (83 percent). All six patients in whom iron therapy was not effective had underlying medical illnesses. Three of the six patients had severe chronic renal insufficiency: two underwent long-term dialysis, and the third had an episode of melena but declined evaluation. Of the three other patients who did not have a response to iron therapy, one subsequently presented with calcinosis, Raynaud's phenomenon, esophageal hypomotility, sclerodactyly, and telangiectasia (CREST syndrome) and reported recurrent epistaxis; one had multiple medical problems, including chronic liver disease; and one with degenerative joint disease later reported having taken nonsteroidal antiinflammatory drugs, and a repeat esophagogastroduodenoscopy showed severe gastritis. Two patients died: one from chronic renal failure and the other (who had responded to iron therapy) from chronic liver disease.
Discussion
Idiopathic iron-deficiency anemia in adults is widely believed to result from chronic colonic blood loss due to mass lesions1,2. However, there are no study-based guidelines for efficiently managing this disorder. In an attempt to establish such guidelines, we prospectively studied 100 patients with iron-deficiency anemia, using upper gastrointestinal endoscopy and colonoscopy followed by small-bowel enteroclysis. We identified lesions consistent with chronic blood loss in 62 of the 100 patients. Although colorectal cancers were found in 11 patients, more lesions were detected in the upper gastrointestinal tract.
Symptoms that could be traced to either the upper or lower gastrointestinal tract were strongly associated with identification of a lesion by esophagogastroduodenoscopy or colonoscopy, respectively. Furthermore, the combination of a positive test for fecal occult blood and site-specific gastrointestinal symptoms increased the likelihood of detecting a bleeding source in the corresponding site. These results strongly suggest that for iron-deficient patients with specific symptoms, the initial examination should be directed by those symptoms. However, asymptomatic patients, particularly those who are elderly, since they have an increased risk of colorectal neoplasia,19 should undergo colonoscopy, followed immediately by an endoscopic examination of the upper gastrointestinal tract if the colonoscopic examination is negative.
For patients in whom a substantial upper gastrointestinal lesion (according to our criteria) has been detected by esophagogastroduodenoscopy, the decision to examine the lower gastrointestinal tract may be controversial. Our data suggest that a synchronous lesion is unlikely. Clinicians may therefore reasonably elect not to undertake a colonic evaluation if a lesion has been detected in the upper gastrointestinal tract. This approach should limit the costs and potential risks of additional studies. Clinical judgment and follow-up, however, are essential. For example, since most of the substantial lesions of the upper gastrointestinal tract are related to acid-peptic disease, an intensive course of anti-acid therapy and iron replacement can be expected to correct the anemia. If the anemia does not resolve or if it returns after therapy has been discontinued, then a reevaluation, including a colonic examination, is recommended.
Nonsteroidal antiinflammatory drugs and alcohol have been associated with acute and chronic injury of the upper gastrointestinal mucosa20,21,22,23,24. We were unable, however, to find any correlation between the use of these irritants and lesions in the upper gastrointestinal tract. The patients who were taking nonsteroidal antiinflammatory drugs may have had ulcers that had healed by the time of our investigation, or the drugs may have induced small gastric or small-bowel lesions that were not detected by endoscopy.
Chronic blood loss from the distal small bowel may be responsible for iron-deficiency anemia1,2,25. Enteroclysis theoretically should reveal substantial mass lesions or severe ulcerative processes but not subtle mucosal abnormalities. In this regard, we may have failed to identify vascular ectasias or subtle abnormalities of the small-bowel mucosa. Although it is possible that some of our patients had occult small-bowel disorders, such as adult celiac disease or Whipple's disease, that were not detected, this possibility seems unlikely. No patients had symptoms of malabsorption, and the majority of patients with negative studies had a response to oral iron therapy. The role of enteroscopy in identifying abnormalities in the small bowel is controversial26. Vascular ectasias and other subtle mucosal lesions have occasionally been detected by this procedure in patients with occult gastrointestinal bleeding or iron-deficiency anemia27,28. Although theoretically appealing, small-bowel enteroscopy is expensive and time consuming, requires special equipment, and has not been rigorously studied.
Over one third of the patients in our study had no lesions compatible with gastrointestinal bleeding. Follow-up data in these patients suggested a favorable prognosis, and the subsequent discovery of gastrointestinal tract lesions was rare. Most of these patients were asymptomatic. Conditions such as unreported menstrual losses (potentially in three patients in our study) or insufficient dietary intake of iron may have been responsible for occult iron deficiency. Alternatively, ulcerations or benign lesions may have caused blood loss in the past but then healed and thus were not detected. Although less likely, small ulcers or polyps may have led to chronic bleeding. Finally, abnormalities such as polyps or vascular ectasias may have been missed during the endoscopic examination. All the patients with persistent anemia had serious underlying medical illnesses, which is consistent with the possibility that their anemia was associated with chronic disease.
We conclude that substantial gastrointestinal lesions, particularly those of the upper gastrointestinal tract, are common in patients with iron-deficiency anemia. Therefore, a thorough evaluation of all such patients is mandatory. Many patients believed at first to be iron-deficient may not have decreased iron stores, and careful screening is therefore important before potentially expensive and unproductive studies are undertaken. We believe that the initial investigation should be directed toward site-specific symptoms. In asymptomatic older patients, the colon should be examined first. Patients with negative colonic studies should then undergo an upper gastrointestinal evaluation. Since synchronous lesions of the upper and lower gastrointestinal tract are rare, once an obvious source of potential blood loss has been identified, further examination of the gastrointestinal tract (but not further follow-up) appears to be unwarranted.
We are indebted to the nurses, staff, residents, and fellows of the gastrointestinal unit at San Francisco General Hospital for their assistance and support, and to Dr. David Heilbron for performing the statistical calculations.
Source Information
From the Gastroenterology Division, Medicine Service, San Francisco General Hospital, University of California, San Francisco. Presented in part at the annual meeting of the American Gastroenterological Association, San Francisco, May 9-15, 1992.
Address reprint requests to Dr. Rockey at the Liver Center Laboratory, San Francisco General Hospital, Bldg. 40, Rm. 4102, San Francisco, CA 94110.
References
The New England Journal of Medicine is owned, published, and copyrighted © 2004 Massachusetts Medical Society. All rights reserved.
 |
 |
 |
 |
