Associated Disorders and Clinical Manifestations
Fabiola Delco, MD, Hashem B. El-Serag, MD, Amnon Sonnenberg, MD

Celiac Sprue affects subjects of Northern European descent more frequently than other ethnic groups, such as Africans, Asians, Japanese or Chinese. Until now only few studies have analyzed the epidemiology of celiac sprue in the United States, where the disease seems to be diagnosed less frequently than in Europe. Its etiology has remained largely unknown; besides an abnormal immune response to dietary gluten, genetic and environmental factors appear to contribute to its development. Multiple associations between celiac sprue and other disorders, in particular autoimmune diseases, have been described. An increased risk of intestinal lymphoma and gastrointestinal neoplasms were also observed in patients with celiac sprue. The clinical manifestations of celiac sprue are only partially of gastrointestinal nature, and a large number of patients present initially with systematic manifestations or with symptoms related to malnutrition.

The clinical epidemiology of celiac sprue is difficult to investigate, because even large gastroenterological centers manage only a few patients with the disease. It is difficult to establish the statistical significance of many associations between celiac sprue and other rare disorders in small patient populations. The largest study dealing with the associations between celiac sprue and other comorbid conditions included less than 350 patients. Most epidemiological studies were carried out in pediatric populations, while the presentation of celiac sprue in adults is less well known. A better understanding of the clinical manifestations of celiac sprue in adults could help the clinician to identify this underdiagnosed disease more frequently.

The Department of Veterans Affairs (VA) maintains computerized records of all patients treated in VA hospitals throughout the United States. The Patient treatment file includes the data of one million hospital discharges per year. Such a large database provides a unique opportunity to study the comorbid occurrences of rare disorders in identical patients. Accordingly, the aim of the present case-control study was to analyze disease associations and clinical manifestations of celiac sprue in an adult population. In relying on a large database, we hoped to establish the epidemiological pattern of celiac sprue with greater confidence than previously possible.

Materials and Methods
Data Source: In the present study, the Patient Treatment File (PTF) of the VA served as the data source. This computerized database contains the records of all US military veterans treated as inpatients in VA hospital throughout the United States. The PTF is managed by the VA Central Automation in Austin, Texas. Each annual file includes the records of about one million hospitalizations among more than 600,000 patients, who can be identified by their unique social security number. In addition to demographic characteristics, each patient’s record contains one primary and up to nine secondary discharge diagnoses. Each discharge diagnosis is encoded according to the 9th revision of the International Classification of Diseases (ICD). Professional coders perform the translation of the diagnoses from the original discharge records into ICD codes.

Identification of Case and Control Subjects: The case subjects constituted all patients with a primary or secondary diagnosis of celiac sprue who were discharged between 1986 and 1995. After identifying the case subjects, the remaining population of each file served for selection of the control subjects. For each case subject, five control subjects were randomly selected from the same annual data file as the case subject. Random choice without replacement was the method employed in the selection procedure so that no control subject could be selected twice for any given year. In a final step, we excluded case and control subjects who had been admitted more than once during consecutive years. If an individual patient appeared in the discharge records more than once, only his/her first appearance was considered. An unequal number of case and control subjects were excluded due to variations in the rate of multiple admissions among case and control subjects. Demographic data regarding gender, ethnicity and age were obtained for case and control subjects alike.

Extraction of Comorbid Conditions: The individual social security numbers were used to follow each case and control subject through the annual computer files. The complete past medical history for all cases and control subjects were obtained by accumulating their ICD codes from all hospital discharges between 1986 and 1995. Duplicates of ICD codes in the same individual were then eliminated. The terms used in the manuscript refer to the original ICD codes developed by the World Health Organization. It should be kept in mind that the terms used to classify comorbid conditions were coined neither by the treating physician nor the authors and that the ICD codes, at times, may provide only crude or even inaccurate descriptions of the underlying clinical syndromes or disease manifestations.

Statistical Analysis: In the univariate analyses, the frequencies of each five-digit ICD code in the case and control population was compared. For comparison, an unadjusted odds ratio and its 95% confidence interval was calculated according to Woolf’s method. This analysis served to identify which diagnoses were significantly associated with celiac sprue. The race and sex distributions of the case and control subjects were compared using chi-squared tests. The mean age of case and control subjects was compared using an unpaired t test.

In the multivariate analysis, all variables, with the exception of age were coded as dichotomous variables, using the code 1 to indicate the presence of the attribute of interest and 0 to indicate its absence. Age, gender, ethnicity and the occurrences of particular diagnoses were chosen as predictor variables. Diagnoses were only considered as predictor variables if they had been identified previously by the univariate analysis. The PROC LOGISTIC command of SAS was used for the multivariate logistic regression. Odds ratios and their 95% confidence intervals served to describe the strength of the influence exerted by each individual predictor. The Wald’s chi-square test tested the significance of each influence. A correlation matrix was calculated to detect and prevent the occurrence of collinearity among the predictor variables.

Results
From 1986 until 1995, 458 individual subjects were diagnosed with celiac sprue. A total of 2692 subjects without the disease were selected as random controls from the entire patient population of 2.67 million discharged from the VA hospitals during the same time period. The demographic characteristics of the case and control group are displayed in Table 1. Patients with celiac sprue were significantly older than control subjects (t=6.265 P<0.0001). Compared with controls, there were relatively more whites than non-whites among patients with celiac sprue (chi-squared=79.453, DF=1, P<0.0001). Although the case population contained relatively more women than the control population, this difference failed to reach statistical significance (chi-square=2.627, DF=1, P=0.105).

The computerized patient records of the Central Automation Center in Austin contain only a selected list of patient characteristics. The condensed patient data are submitted electronically to the Automation Center from the VA hospitals; however, the actual records are kept at the 172 individual facilities throughout the United States. It is therefore not possible to review all patient records of the entire case population for accuracy and means of diagnosis. Nevertheless, to obtain an idea about the general validity of the diagnosis, we selected seven records from patients diagnosed with celiac sprue at the two VA hospitals in Albuquerque, New Mexico and Milwaukee, Wisconsin. In all seven cases, the diagnosis of celiac sprue had been based on multiple mucosal biopsies. The chart review also confirmed the correctness of 23 out of 26 secondary discharge diagnoses. No instance of gross discrepancy between the chart diagnoses and the computerized ICD code was noted.

Table 1: Demographic Characteristics of Case and Control Subjects

The second table contains the list of diseases found to be significantly associated with celiac sprue according to the results of the univariate analyses. The table compares the prevalence rates of systemic, gastrointestinal, and nutritional manifestations in case and control subjects. In the group of systemic and gastrointestinal manifestations, dermatitis herpetiformis, pancreatic insufficiency, and lactase deficiency represented the most frequently diagnosed comorbid diseases. All comorbid conditions were relatively infrequent, as none of the prevalence rates exceeded 7%. Weight loss, cachexia and nutritional marasmus represented general nutritional manifestations while various forms mineral and vitamin deficiencies constituted the specific symptoms. The most common clinical symptom of malnutrition was iron-deficiency anemia. In patients with celiac sprue, the prevalence weight loss, cachexia, hypocalcemia, osteopenia, vitamin B-complex deficiency and vitamin B12-deficiency anemia ranged between 5% and 7%, compared to a range of only 1-2% in the control population.

Table 2: Systemic, gastrointestinal, and nutritional manifestations in case and control subjects.

The third table lists the odds and ratios and the confidence intervals for the comorbid associations of celiac sprue. The odds ratios and their corresponding confidence intervals were obtained through multivariate logistic regression. The upper half of the table illustrates the spectrum of systemic and gastrointestinal diagnoses. The systemic manifestations include enlargement of lymph nodes, malignant lymphomas, dermatitis herpetiformis and pulmonary eosinophilia. The gastrointestinal conditions include functional bowel symptoms, lactase deficiency, regional enteritis, pancreatic insufficiency and chronic active nonalcoholic hepatitis. In the multivariate analysis no significant association between celiac sprue and gastrointestinal cancers was found. This lack of association concerned all gastrointestinal cancers, as well as individual cancer types, such as esophageal, gastric or colorectal cancer. However, the univariate analysis showed a trend in favor of an increased prevalence rate of gastrointestinal cancers in celiac patients (7.25% vs. 4.1%). The multivariate analysis also failed to show a significant association between celiac sprue and ulcerative colitis, although the univariate analysis suggested a trend in favor of such an association (1.1% vs. .4%). In the univariate analyses, individual forms of insulin-dependent diabetes mellitus (IDDM) were found to be associated with a significant odds ratio; for instance, IDDM with ophthalmic manifestations, and neurological manifestations and with other unspecified complications. However, in the multivariate analysis, individual codes or a summary of all IDDM-related ICD codes failed to reach a statistical significance.

Table 3: Multivariate analysis, nutritional manifestations of celiac sprue.

All nutritional manifestations revealed by the univariate analysis were found to also be statistically significant in the multivariate analysis. Nutritional marasmus, cachexia, weight loss, hypocalcemia, osteopenia, vitamin B-complex deficiency, Iron deficiency and vitamin B12-deficiency anemia were 2.4-5.6 times more common in patients with celiac sprue than in control subjects. Although the anemias of a combined folate-vitamin B12or a folate deficiency alone affected only 2-3% of the cases, in the multivariate logistic regression these two anemias were associated with the highest odds ratio.

Discussion
This is one of the largest studies to examine the prevalence of comorbid diseases in adult patients with celiac sprue. The previously reported increased risk for malignant lymphoma was confirmed by the present results yielding an odds ratio of 4.5. Besides corroborating well-known associations between celiac sprue and its related conditions, such as enlargement of lymph nodes, dermatitis herpetiformis, lactase deficiency and pancreatic insufficiency. Less documented associations also became evident by our analysis. Pulmonary eosinophilia, chronic active nonalcoholic hepatitis and Crohn’s disease belong in this group of diseases, for which an association with celiac sprue was previously suggested but never statistically proven. The nutritional consequences of celiac sprue could be divided into two large subgroups involving general and specific manifestations. The general manifestations, such as weight loss, cachexia and nutritional marasmus, occurred 2.4-4.5 times more in celiac patients than in controls. The specific manifestations comprised of iron, calcium and vitamin deficiencies and affected celiac patients 2.6-5.6 times more often than controls.

Celiac sprue is an infrequently diagnosed disease among North Americans as compared with Northern Europeans and a large database is needed to recruit an appreciable number of celiac patients. Besides providing such an opportunity, the Patient Treatment File of the VA offers the additional advantage of investigating celiac sprue is a unique group of patients. Different from most previous studies, our population is comprised exclusively of adults. The natural history and the presentation of celiac sprue in the elderly is less well characterized than in children or adolescents. The natural history and the long-term sequence of celiac sprue may become more readily discernible in a patient population of adults with long-standing disease. However, the occurrence of comorbid diseases revealed by the present analysis of mostly adult males may also differ from the ones of previous studies, which included large fractions of women and children with the new-onset disease.

The possibility to compare the occurrence of comorbid diseases among patients with celiac sprue and a control group of subjects without this disease represents an important advantage of working with a large computerized database. In addition to a description of comorbidity and its prevalence rate, one can subject such patterns to statistical analyses. However, the reliance on patient records retrieved from a database also has a downside. Patients with celiac sprue managed entirely, as outpatients would not appear in our database. Unlike a prospective study, we could not ascertain uniform standards of diagnosis among patients distributed throughout the United States. Many of the clinical symptoms associated with celiac sprue, such as weight loss, diarrhea or abdominal cramps may not have been recorded as a discharge diagnosis. The dependence on discharge diagnoses alone as an identification method may result in the underrepresentation of comorbid disease may have gone unrecognized or may not have been considered irrelevant in the management of the acute celiac disease. The computerized patient record may capture only a fragmented and incomplete picture of a complex disease with involvement of multiple organs. Various forms of anemia, chronic lung disease, hepatitis, and pancreatitis are frequent causes for hospital admission among veterans. For these reasons, our odds ratios may have been subjected to various forms of detection bias.

Weight loss, cachexia and nutritional marasmus were diagnosed in 5-7% of the patients with celiac sprue as compared to 1-2% in the control group. These prevalence rates of malnutrition in celiac patients are lower than reported in previous studies. The difference may be related to the older age of the present patient population. Various forms of anemia represented the most frequent nutritional manifestations of celiac sprue. Similar results were observed in the above-cited studies, with prevalence rates of anemia ranging between 22% and 90%. We found an increased risk for iron, vitamin B12, folate, and combined deficiency anemias in celiac patients. Celiac sprue affects primarily the proximal parts of the small intestine, while vitamin B12 is absorbed in the terminal ileum. Therefore mechanisms other than villous atrophy must be responsible for its deficiency. Previous authors demonstrated a lymphocytic infiltration of the gastric and small and large intestinal mucosa in celiac sprue. The occasional bacterial overgrowth of the small intestine in untreated celiac sprue could reduce the amount of intraluminal vitamin B12 available for absorption. Iron-deficiency anemia was the most frequently occurring anemia; mucosal malabsorption of iron by the duodenal mucosa is likely to be the most important mechanism. Existing hypochlorhydria secondary to gastric lymphocytic infiltration could also interfere with iron absorption.

We confirmed the increased risk of developing malignant lymphoma found in other populations. Lymphoma occurred in 3% of our case population. Higher prevalence rates of lymphoma were reported by previous studies. However, the diagnosis was frequently established by postmortem examination, and the enrollment of patients into the study was biased towards patients with lymphoma. Lymphoma is commonly underdiagnosed during life, because symptoms of therapy resistant celiac sprue can mask those of lymphoma. Superficial lymphadenopathy was found at initial presentation in 41% of the celiac patients with lymphoma, the lymphadenopathy itself tended to appear rather late in the course of the disease. However, no data exists about the predictive value of peripheral lymphadenopathy with respect to the development of malignant lymphoma.

Since 1970, there have been multiple reports of various bronchopulmonary diseases coexisting with celiac sprue. In most instances of such comorbid occurrences, a diagnosis of bird fanciers lung, cryptogenic fibrosing alveolitis, interstitial lung disease, or asthma was made. However, occasionally the comorbid lung disease was classified as farmer’s lung, or idiopathic pulmonary hemosiderosis. Our comparison of celiac patients with controls revealed an increased occurrence of the ICD code 518.3 in the celiac population. This code refers to pulmonary eosinophilia, asthma, and Loffler syndrome. The majority of authors agree on the fact that a portion of celiac patients presents with symptoms suggestive of an immune-mediated lung disease. The large variety of medical terms used to describe this entity appears to reflect a limitation of the current coding system in capturing this peculiar pulmonary disease. Accordingly, the terms pulmonary eosinphilia or eosinophilic asthma, used in the present study, may not represent the exact medical conditions that coexisted with celiac sprue, but the physicians effort to specify the lung disease of this patient population.

Previous studies examined the exocrine pancreatic function in children and adults with celiac sprue in an attempt to clarify its mechanisms. In a study of 111 children with celiac sprue, Carroccio et al demonstrated that pancreatic dysfunction did not occur if dietary restriction of gluten restored the normal appearance of intestinal mucosa. We also found an increased prevalence of chronic active nonalcoholic hepatitis in two case reports, Kahn et al and Bjorneklett et al; a celiac patient was described with having granulomatous hepatitis. In the report by Bjorneklett, the granulomas disappeared after dietary treatment. Two other studies alluded to an association between celiac disease and sarcoidosis. It appears that in a fraction of patients, their celiac disease results in a form of active hepatitis that has remained unclassified and whose etiology is poorly understood. Our study also shows a significant association between celiac sprue and Crohn’s disease. In 1962, Shiner and Drury reported partial and subtotal villous atrophy in 6 out of 11 patients with Crohn’s disease. By contrast, Milewski et al could not find any villous atrophy in a group of 41 patients with regional enteritis. Subsequently, multiple case reports described the coexistence of celiac sprue with inflammatory bowel diseases, but none of these studies was able to provide statistical evidence for this association.

In summary, the present case-control study describes the systemic and nutritional manifestations of celiac sprue in adult patients with long-standing disease. Our study corroborates the known associations between celiac sprue and dermatitis herpetiformis, lactase deficiency, and enlargement of lymph nodes and lymphoma. Moreover, celiac sprue in adults is associated with various systemic manifestations, such as hepatitis, regional enteritis, pancreatic insufficiency and pulmonary eosinphilia. This variety of complex associations points at a systemic disease that involves multiple organs and exceeds an isolated nutritional intolerance.

*Source: Digestive Diseases and Science, Vol. 44, No. 5 May 1999. Pages 966-972.

**Permission for reprinting in Lifeline granted through Dr. Amnon Sonnenberg, gastroenterology section, Department of Veterans Affairs, U.S. Veterans Medical Center, Albuquerque, New Mexico.