Empirically Incorrect
http://www.100md.com
《新英格兰医药杂志》
In this Journal feature, information about a real patient is presented in stages (boldface type) to an expert clinician, who responds to the information, sharing his or her reasoning with the reader (regular type). The authors' commentary follows.
A 46-year-old Mexican immigrant presented to the hospital because of epigastric pain and the vomiting of coffee-grounds material. He had had fatigue, malaise, jaundice, and a 20-lb (9.1 kg) weight loss during the previous two months. He had also had dark stools, light-headedness, and mild shortness of breath, but no fever, chills, or night sweats.
The patient had upper gastrointestinal bleeding and shortness of breath and lightheadedness that can be attributed to anemia from blood loss. The major causes of upper gastrointestinal bleeding are duodenal ulcer, gastric erosions or ulcer, and gastroesophageal varices. Weight loss is not to be expected in a patient with duodenal ulcer and therefore suggests a chronic inflammatory condition or cancer. Hispanic populations and other immigrant groups in the United States are disproportionately affected by Helicobacter pylori infection. Infection with intestinal parasites should also be considered in persons who have emigrated from areas where such infections are endemic. Nonhepatic causes of jaundice, such as hemolysis, are less common than hepatobiliary disease, although the character of the bleeding in this case is not consistent with varices. A patient with cholecystoduodenal fistula, most commonly associated with chronic cholecystitis, may present with the combination of jaundice and upper gastrointestinal bleeding. Hemobilia, another uncommon diagnosis, classically manifests with the triad of melena, biliary colic, and jaundice. It is also possible that there are multiple causes for this patient's symptoms.
The patient worked on livestock farms in Mexico before moving to the United States 15 years earlier. His medical history was unremarkable, and he said he did not use alcohol or illicit drugs. He was not taking any medications or herbs. His most recent travel had been to Mexico three years previously. He was married with four children, worked as a gardener, and had no pets.
The history of exposure to livestock leads me to consider diagnoses such as brucellosis and infection with Echinococcus granulosus and Fasciola hepatica, which are common in regions where sheep and cattle are raised. If the patient had had brucellosis, the latent period would not have been so long; in contrast, a patient infected with E. granulosus is frequently asymptomatic until the hepatic cysts become larger than 10 cm. Such cysts cause pressure or mass effects that can result in obstructive jaundice and portal hypertension. Jaundice is a classic manifestation of chronic infestation of the biliary and hepatic ducts by the F. hepatica fluke and may lead to secondary biliary cirrhosis and portal hypertension.
On physical examination, the patient had a temperature of 37.7°C, a pulse of 106 beats per minute, a blood pressure of 127/54 mm Hg, and 18 respirations per minute. His oxygen saturation was 100 percent while he was breathing ambient air. He had obvious jaundice, without palpable lymphadenopathy. His cardiac examination was normal, except for tachycardia. His lungs were clear to auscultation and percussion. There was mild epigastric tenderness without hepatosplenomegaly or masses. He had no signs or symptoms of chronic liver disease. The neurologic examination was normal. The rectal examination revealed melena.
Given the results of the rectal examination and the tachycardia, the patient should be assessed for postural hypotension. Routine laboratory tests, including the measurement of serum total bilirubin, should be followed by ultrasonography of the right upper abdominal quadrant.
Laboratory evaluation revealed a white-cell count of 3300 per cubic millimeter; a hemoglobin level of 4.5 g per deciliter, with a mean corpuscular volume of 87.4 μm3; and a platelet count of 361,000 per cubic millimeter. Analysis of a peripheral-blood smear showed 13 percent band forms, 27 percent monocytes, and 25 percent nucleated red cells, with evidence of basophilic stippling but no schistocytes. Levels of electrolytes were normal, except the blood urea nitrogen level was 36 mg per deciliter (12.9 mmol per liter). Serum chemical analysis revealed the following values: albumin, 2.7 g per deciliter; alkaline phosphatase, 170 U per liter (normal range, 53 to 128); alanine aminotransferase, 25 U per liter (normal range, 10 to 35); aspartate aminotransferase, 36 U per liter (normal range, 14 to 50); and total bilirubin, 12.7 mg per deciliter (217 μmol per liter; normal range, 0.3 to 1.2 ), with a direct bilirubin level of 7.7 mg per deciliter (131.67 μmol per liter; normal, less than 0.2 ). The level of lactate dehydrogenase was 210 U per liter (normal range, 100 to 190), and the haptoglobin level was 33 mg per deciliter (normal range, 33 to 255). The international normalized ratio was 3.4 before the administration of vitamin K and subsequently returned to normal. A radiograph of the chest showed no abnormalities. Computed tomography (CT) of the abdomen showed a nodule at the base of the right lung and lymphadenopathy in the upper portion of the abdomen and retroperitoneum. Endoscopy revealed a duodenal ulcer; a urease enzyme test of a biopsy specimen was positive for H. pylori.
Abdominal pain, anemia, and basophilic stippling can be manifestations of lead poisoning, but the neuropsychiatric manifestations of this disease are absent in this patient. Drinking water in Mexico can be contaminated with arsenic, and chronic exposure has been associated with cholestatic liver disease, hepatic angiosarcomas, and cancers of the lung and bladder. However, the patient emigrated from Mexico 15 years previously, and no skin lesions were found on presentation. A duodenal ulcer associated with H. pylori may cause epigastric pain and anemia, but it would not explain the jaundice or hematologic abnormalities, including the findings on the peripheral-blood smear. Nucleated red cells in the absence of hemolysis are consistent with the presence of a myelophthisic process, such as metastatic cancer or infection. Thus, a bone marrow biopsy should be performed, particularly in view of the presence of the lung nodule and abdominal lymphadenopathy. The absence of biliary dilatation on abdominal CT is consistent with the presence of intrahepatic cholestasis. Viral or toxic hepatitis, primary sclerosing cholangitis, or drugs are the most likely cause of the cholestasis in this case. Paraneoplastic syndromes associated with cancer, particularly renal-cell carcinoma and malignant lymphoproliferative disorders, may occasionally lead to cholestatic liver disease. A liver biopsy would probably be required for a further characterization of the disease process.
The patient was a Jehovah's Witness; therefore, his treatment was limited to iron dextran, folic acid, and erythropoietin. A fever (temperature of up to 38.8°C) developed on the third hospital day, which subsequently resolved without treatment. A test for human immunodeficiency virus and a tuberculin test (without controls) were negative. A biopsy specimen of the bone marrow was negative for a malignant tumor or granulomas and showed adequate iron stores. Diagnostic laparoscopy revealed a grossly congested liver with bilious ascites and diffuse lymphadenopathy of the lesser omentum. A lymph node obtained from the lesser curvature of the stomach contained caseating granulomas; a simultaneously obtained liver specimen was characterized by marked hepatocellular and canalicular cholestasis with mild ductopenia, but no fibrosis or cirrhosis. Staining of all tissues for acid-fast bacilli was negative. On the basis of the insidious nature of this patient's illness, the fever and lymphadenopathy, and the history of exposure to livestock, he was treated empirically for brucellosis with doxycycline and gentamicin; in addition, he was treated for H. pylori with amoxicillin, clarithromycin, and omeprazole. Despite the lack of a clear diagnosis, the patient was discharged to his home with a plan for close follow-up.
The normal bone marrow in the presence of circulating nucleated red cells and jaundice raises the possibility of a life-threatening disorder, such as sepsis. Brucellosis has protean manifestations, the most prominent of which is fever. Mild elevations in aminotransferases may occur, but the degree of hyperbilirubinemia seen in this case would be atypical. The caseating granulomas suggest the presence of tuberculosis, although caseation is not specific for mycobacterial infection; fungal and atypical bacterial infections should also be considered. Although it is not unreasonable to initiate empirical antibiotic therapy for suspected brucellosis, I am concerned about discharging the patient without a diagnosis, particularly since patients with hyperbilirubinemia and circulating nucleated red cells have a high in-hospital mortality rate.
At a follow-up visit one week later, the patient had no improvement in his symptoms. The white-cell count was 1100 per cubic millimeter, with an absolute neutrophil count of 200. The hemoglobin level was 11.5 g per deciliter; the platelet count was normal. Serologic tests for brucella and viral hepatitis were negative, as were tests for antinuclear, antineutrophil cytoplasmic, anti–smooth-muscle, and antimitochondrial antibodies. He was readmitted to the hospital. The fall in the white-cell count was attributed to the antibiotics, and it resolved after treatment with granulocyte colony-stimulating factor (G-CSF) and the discontinuation of his medications. He remained afebrile. Magnetic resonance cholangiopancreatography showed no dilatation of pancreatic or intrahepatic biliary ducts. Empirical, four-drug therapy for tuberculosis was begun, and the patient was discharged.
The neutropenia may have been a manifestation of worsening liver disease. Although the neutropenia resolved with the discontinuation of the antibiotics and the administration of G-CSF, the antibiotics that were given are not commonly associated with this adverse event. Of the four antibiotics, doxycycline and gentamicin are the least likely to be the offending agents. Brucellosis and tuberculosis are both associated with neutropenia. Although Brucella melitensis is the species that most commonly infects humans, infection with B. canis can also occur and does not result in antibodies against standard brucella antigens. Serum agglutination tests may also be negative early in the course of disease, necessitating repeated testing. Cultures for tuberculosis may be negative in the presence of miliary tuberculosis, but committing the patient to a six-month course of four-drug therapy for tuberculosis seems premature at this point. Brucella species are slow-growing organisms, and it may take as long as several weeks for the cultures to become positive. I would reinitiate treatment with doxycycline and gentamicin while awaiting the culture results.
The patient was readmitted two weeks later because of a white-cell count of 1700 per cubic millimeter and an absolute neutrophil count of 750. The level of alanine aminotransferase was 31 U per liter; that of aspartate aminotransferase, 41 U per liter; that of alkaline phosphatase, 350 U per liter; and that of total bilirubin, 11.9 mg per deciliter (203 μmol per liter). Repeated bone-marrow aspiration revealed no abnormalities. CT of the chest and abdomen showed enlarging retroperitoneal lymph nodes with new lymph nodes above the right main bronchus and in the left inguinal region (Figure 1). Surgical biopsy specimens of the hilar and inguinal lymph nodes revealed reactive lymphoid tissue and no granulomas; staining for acid-fast bacilli was negative. The neutropenia resolved after discontinuation of the antituberculosis treatment. The patient was discharged while taking empirical prednisone (60 mg daily orally) for presumed granulomatous hepatitis.
Figure 1. CT Scans of the Chest (Panel A) and Abdomen (Panel B).
Panel A shows a lymph node that is 2.5 cm in diameter and superior to the right main bronchus (arrow). Panel B shows a lymph node that is 2.4 cm in diameter and adjacent to the aorta and inferior vena cava (arrow).
The level of alkaline phosphatase has doubled since the initial laboratory evaluation, consistent with the presence of an infiltrative or cholestatic process. Granulomatous hepatitis is a syndrome characterized by prolonged recurrent fevers, myalgias, arthralgias, and granulomas in the liver and other organs. However, granulomatous hepatitis is a diagnosis of exclusion, and before empirical corticosteroid therapy is instituted, four to eight weeks of antituberculosis therapy is typically recommended. In the absence of a clear-cut diagnosis, I am concerned about the use of empirical therapy, particularly corticosteroids, given that infectious causes have not been ruled out. Instead, I would recommend repeating the liver biopsy, with cultures.
After vacationing in Mexico for a month, during which he continued corticosteroid therapy, the patient returned to the clinic with persistent weight loss and abdominal pain, but without fever. He was readmitted to the hospital and was found to have an alanine aminotransferase level of 818 U per liter, an aspartate aminotransferase level of 379 U per liter, an alkaline phosphatase level of 2132 U per liter, and a total bilirubin level of 36 mg per deciliter (616 μmol per liter). Endoscopic retrograde cholangiopancreatography showed relatively thin intrahepatic ducts, with a decreased number of ducts in the left hepatic lobe. He was transferred to a liver-transplantation center, where examination of a specimen from the second liver biopsy showed marked hepatocellular and canalicular cholestasis and no intact interlobular bile ducts.
This patient has a rapidly progressive intrahepatic cholestatic process. Drugs such as phenothiazines and environmental exposure to arsenic have been implicated in a syndrome characterized by progressive ductopenia: the vanishing bile-duct syndrome. However, on initial presentation the patient was not taking medications. Arsenic poisoning can result from exposure to contaminated well water or from the use of or exposure to arsenic-containing pesticides and herbicides and pressure-treated wood; as a gardener, he may have been exposed to these substances. Another herbicide, paraquat, has been associated with a cholangiodestructive process, although most cases of paraquat poisoning are the result of suicide attempts rather than accidental exposure. Lymphoma-associated ductopenia and idiopathic adulthood ductopenia have been described, including a variant of the latter that is characterized by symptoms of decompensated cirrhosis at presentation and a paucity of intrahepatic bile ducts on liver biopsy and that frequently requires orthotopic liver transplantation. The administration of corticosteroids might have been expected to improve a lymphoproliferative process, if only transiently. I admit that I remain puzzled. If I were caring for this patient, I would discontinue the prednisone and restart the antituberculosis therapy while contemplating what other types of tissue I should obtain to make a diagnosis.
After a progressive downhill course complicated by diffuse pulmonary infiltrates, clinical sepsis, and two cardiopulmonary arrests, the patient died. At autopsy, he was found to have had Hodgkin's disease in the mediastinal and mesenteric lymph nodes (Figure 2) and marked cholestasis and ductopenia in the liver (Figure 3). A suppurative, necrotizing fungal pneumonia was present throughout the lungs. There was no identifiable pulmonary malignant tumor or evidence of hepatic fungal infection. A diagnosis of vanishing bile-duct syndrome due to Hodgkin's disease was made. A retrospective review of all previous biopsy specimens showed no evidence of Hodgkin's disease.
Figure 2. Specimen of an Abdominal Lymph Node Obtained at Autopsy, Showing Infiltrates of Lymphocytes, Plasma Cells, Eosinophils, and Reed–Sternberg Cells (Arrow).
Figure 3. Biopsy Specimen of the Liver Obtained from the Patient at Autopsy (Panel A) and Biopsy Specimen of Normal Liver (Panel B) (Hematoxylin and Eosin).
Bile ducts are absent in the portal tract shown in Panel A, whereas a bile duct is present in the portal tract shown in Panel B (arrow).
Cholestasis can occur in patients with Hodgkin's disease because of extrahepatic biliary obstruction from extrinsic compression by enlarged lymph nodes, direct hepatic infiltration, or a rare but well-recognized paraneoplastic process that is characterized by intrahepatic cholestasis, with or without bile-duct paucity. In retrospect, what is striking about this case is the lack of evidence of Hodgkin's disease on the antemortem lymph-node examination. Clearly, a diagnosis of Hodgkin's disease should be considered in any patient with unexplained intrahepatic cholestasis, particularly a febrile cholestatic illness. Perhaps this case best illustrates the problem of relying too much on too little information with which to base empirical therapy.
Commentary
Vanishing bile-duct syndrome is a rare entity characterized by the progressive loss of small intrahepatic bile ducts, leading to cholestatic hepatitis and often death. Ductopenia may be due to immunologic, infective, vascular, drug-induced, paraneoplastic, or idiopathic causes.1 Since cholangiography typically shows no abnormalities and the histologic loss of bile ducts may be subtle, the lesion is often missed on initial pathological review.1
The association of vanishing bile-duct syndrome with Hodgkin's disease has been well established and is most likely a paraneoplastic phenomenon.1,2,3,4,5 In this syndrome, there is no evidence of hemolysis, extrinsic compression of the extrahepatic ducts, or tumor infiltration on histologic examination of the liver. Unlike other causes of vanishing bile-duct syndrome, those that occur in association with Hodgkin's disease may be reversible if chemotherapy induces a remission of the cancer.2 Cholestatic hepatitis has been described in several case reports as appearing several weeks before the Hodgkin's disease became clinically apparent.2,6 Although Hodgkin's disease reportedly has remained undiagnosed after an initial bone marrow biopsy,2 we are not aware of reports of a similar difficulty in detecting Hodgkin's disease after numerous lymph-node and bone marrow biopsies, as in this case. This delay in diagnosis, compounded by the subtlety of the bile-duct loss on the initial liver-biopsy specimen, resulted in a substantial delay in identifying the cause of the patient's jaundice.
When a patient's condition worsens despite the use of empirical therapy, clinicians must decide whether to continue the empirical therapy, change the empirical therapy, or repeat diagnostic testing. Such decisions, which must often be made in practice, ultimately require clinical judgment to determine whether the presumed diagnosis is correct and to evaluate the risks of treating or not treating the patient. In this case, the discussant was concerned about the empirical use of corticosteroids. The use of that therapy, together with certain other factors (including the patient's unrecognized underlying lymphoma, the frequent hospitalizations, and the use of intravenous catheters and antibiotics), probably contributed to the ultimate development of fungal pneumonia.
Because a mathematical decision-tree analysis is too complex for physicians' daily use in establishing the best course of action, they tend to make decisions in an incremental or consecutive fashion.7 This approach allows them to use whatever limited data are available, including personal experience and the patient's response to therapy, to decide the next treatment. In this way, each course of failed empirical therapy serves as a new piece of diagnostic information and may help exclude another disorder from the differential diagnosis.
However, when using a consecutive approach to problem solving, clinicians may fail to reanalyze the data on which previous decisions were made. In this case, the lack of an identifiable malignant tumor in numerous biopsy specimens gave the clinicians the false sense that cancer had definitively been ruled out. Thus, infectious and inflammatory conditions became the two most likely categories of disease, leading to a series of empirical treatments aimed at the most probable cause. Ideally, clinicians should retrospectively evaluate the clinical data on which they have based their previous decisions and consider repeating diagnostic studies if the trial-and-error approach seems to be failing.
In this case, the histologic identification of vanishing bile-duct syndrome and the recognition of the potentially latent appearance of Hodgkin's disease in association with this syndrome might have led to earlier diagnosis and potentially lifesaving treatment. This case reminds us of the link between the vanishing bile-duct syndrome and Hodgkin's disease and the frustration and consequences that result from being empirically incorrect.
Supported by a Career Development Award from the Health Services Research and Development Program of the Department of Veterans Affairs (to Dr. Saint) and a Patient Safety Developmental Center Grant (P20-HS11540, to Dr. Saint) from the Agency for Healthcare Research and Quality.
No potential conflict of interest relevant to this article was reported.
We are indebted to Donald Houghton, M.D., Sandra White, M.D., and Tracy Funk of Oregon Health and Sciences University Department of Pathology for reviewing all the biopsy and autopsy specimens pertaining to this case.
Source Information
From the Department of Medicine, Legacy Emanuel and Legacy Good Samaritan Hospitals, Portland, Oreg. (A.S., D.J.G.); and the Department of Medicine, Ann Arbor Veterans Affairs Medical Center (S.S., R.H.M.); Patient Safety Enhancement Program, Ann Arbor Veterans Affairs Medical Center and University of Michigan Health System (S.S.); and the Department of Internal Medicine, University of Michigan Medical School (S.S., R.H.M.) — all in Ann Arbor.
Address reprint requests to Dr. Schmitt at the Legacy Emanuel Hospital and Health Center, 2801 N. Gantenbein Avenue, Suite 4100, Portland, OR 97227, or at aschmitt@lhs.org.
References
Hubscher SG, Lumley MA, Elias E. Vanishing bile duct syndrome: a possible mechanism for intrahepatic cholestasis in Hodgkin's lymphoma. Hepatology 1993;17:70-77.
Crosbie OM, Crown JP, Nolan NPM, Murray R, Hegarty JE. Resolution of paraneoplastic bile duct paucity following successful treatment of Hodgkin's disease. Hepatology 1997;26:5-8.
Birrer MJ, Young RC. Differential diagnosis of jaundice in lymphoma patients. Semin Liver Dis 1987;7:269-277.
Warner AS, Whitcomb FF. Extrahepatic Hodgkin's disease and cholestasis. Am J Gastroenterol 1994;89:940-941.
Jansen PL, van der Lelie H. Intrahepatic cholestasis and biliary cirrhosis associated with Hodgkin's disease. Neth J Med 1994;44:99-102.
Lieberman DA. Intrahepatic cholestasis due to Hodgkin's disease: an elusive diagnosis. J Clin Gastroenterol 1986;8:304-307.
Moskowitz AJ, Kuipers BJ, Kassirer JP. Dealing with uncertainty, risks, and tradeoffs in clinical decisions: a cognitive science approach. Ann Intern Med 1988;108:435-449.(Amy Schmitt, M.D., Daniel)
A 46-year-old Mexican immigrant presented to the hospital because of epigastric pain and the vomiting of coffee-grounds material. He had had fatigue, malaise, jaundice, and a 20-lb (9.1 kg) weight loss during the previous two months. He had also had dark stools, light-headedness, and mild shortness of breath, but no fever, chills, or night sweats.
The patient had upper gastrointestinal bleeding and shortness of breath and lightheadedness that can be attributed to anemia from blood loss. The major causes of upper gastrointestinal bleeding are duodenal ulcer, gastric erosions or ulcer, and gastroesophageal varices. Weight loss is not to be expected in a patient with duodenal ulcer and therefore suggests a chronic inflammatory condition or cancer. Hispanic populations and other immigrant groups in the United States are disproportionately affected by Helicobacter pylori infection. Infection with intestinal parasites should also be considered in persons who have emigrated from areas where such infections are endemic. Nonhepatic causes of jaundice, such as hemolysis, are less common than hepatobiliary disease, although the character of the bleeding in this case is not consistent with varices. A patient with cholecystoduodenal fistula, most commonly associated with chronic cholecystitis, may present with the combination of jaundice and upper gastrointestinal bleeding. Hemobilia, another uncommon diagnosis, classically manifests with the triad of melena, biliary colic, and jaundice. It is also possible that there are multiple causes for this patient's symptoms.
The patient worked on livestock farms in Mexico before moving to the United States 15 years earlier. His medical history was unremarkable, and he said he did not use alcohol or illicit drugs. He was not taking any medications or herbs. His most recent travel had been to Mexico three years previously. He was married with four children, worked as a gardener, and had no pets.
The history of exposure to livestock leads me to consider diagnoses such as brucellosis and infection with Echinococcus granulosus and Fasciola hepatica, which are common in regions where sheep and cattle are raised. If the patient had had brucellosis, the latent period would not have been so long; in contrast, a patient infected with E. granulosus is frequently asymptomatic until the hepatic cysts become larger than 10 cm. Such cysts cause pressure or mass effects that can result in obstructive jaundice and portal hypertension. Jaundice is a classic manifestation of chronic infestation of the biliary and hepatic ducts by the F. hepatica fluke and may lead to secondary biliary cirrhosis and portal hypertension.
On physical examination, the patient had a temperature of 37.7°C, a pulse of 106 beats per minute, a blood pressure of 127/54 mm Hg, and 18 respirations per minute. His oxygen saturation was 100 percent while he was breathing ambient air. He had obvious jaundice, without palpable lymphadenopathy. His cardiac examination was normal, except for tachycardia. His lungs were clear to auscultation and percussion. There was mild epigastric tenderness without hepatosplenomegaly or masses. He had no signs or symptoms of chronic liver disease. The neurologic examination was normal. The rectal examination revealed melena.
Given the results of the rectal examination and the tachycardia, the patient should be assessed for postural hypotension. Routine laboratory tests, including the measurement of serum total bilirubin, should be followed by ultrasonography of the right upper abdominal quadrant.
Laboratory evaluation revealed a white-cell count of 3300 per cubic millimeter; a hemoglobin level of 4.5 g per deciliter, with a mean corpuscular volume of 87.4 μm3; and a platelet count of 361,000 per cubic millimeter. Analysis of a peripheral-blood smear showed 13 percent band forms, 27 percent monocytes, and 25 percent nucleated red cells, with evidence of basophilic stippling but no schistocytes. Levels of electrolytes were normal, except the blood urea nitrogen level was 36 mg per deciliter (12.9 mmol per liter). Serum chemical analysis revealed the following values: albumin, 2.7 g per deciliter; alkaline phosphatase, 170 U per liter (normal range, 53 to 128); alanine aminotransferase, 25 U per liter (normal range, 10 to 35); aspartate aminotransferase, 36 U per liter (normal range, 14 to 50); and total bilirubin, 12.7 mg per deciliter (217 μmol per liter; normal range, 0.3 to 1.2 ), with a direct bilirubin level of 7.7 mg per deciliter (131.67 μmol per liter; normal, less than 0.2 ). The level of lactate dehydrogenase was 210 U per liter (normal range, 100 to 190), and the haptoglobin level was 33 mg per deciliter (normal range, 33 to 255). The international normalized ratio was 3.4 before the administration of vitamin K and subsequently returned to normal. A radiograph of the chest showed no abnormalities. Computed tomography (CT) of the abdomen showed a nodule at the base of the right lung and lymphadenopathy in the upper portion of the abdomen and retroperitoneum. Endoscopy revealed a duodenal ulcer; a urease enzyme test of a biopsy specimen was positive for H. pylori.
Abdominal pain, anemia, and basophilic stippling can be manifestations of lead poisoning, but the neuropsychiatric manifestations of this disease are absent in this patient. Drinking water in Mexico can be contaminated with arsenic, and chronic exposure has been associated with cholestatic liver disease, hepatic angiosarcomas, and cancers of the lung and bladder. However, the patient emigrated from Mexico 15 years previously, and no skin lesions were found on presentation. A duodenal ulcer associated with H. pylori may cause epigastric pain and anemia, but it would not explain the jaundice or hematologic abnormalities, including the findings on the peripheral-blood smear. Nucleated red cells in the absence of hemolysis are consistent with the presence of a myelophthisic process, such as metastatic cancer or infection. Thus, a bone marrow biopsy should be performed, particularly in view of the presence of the lung nodule and abdominal lymphadenopathy. The absence of biliary dilatation on abdominal CT is consistent with the presence of intrahepatic cholestasis. Viral or toxic hepatitis, primary sclerosing cholangitis, or drugs are the most likely cause of the cholestasis in this case. Paraneoplastic syndromes associated with cancer, particularly renal-cell carcinoma and malignant lymphoproliferative disorders, may occasionally lead to cholestatic liver disease. A liver biopsy would probably be required for a further characterization of the disease process.
The patient was a Jehovah's Witness; therefore, his treatment was limited to iron dextran, folic acid, and erythropoietin. A fever (temperature of up to 38.8°C) developed on the third hospital day, which subsequently resolved without treatment. A test for human immunodeficiency virus and a tuberculin test (without controls) were negative. A biopsy specimen of the bone marrow was negative for a malignant tumor or granulomas and showed adequate iron stores. Diagnostic laparoscopy revealed a grossly congested liver with bilious ascites and diffuse lymphadenopathy of the lesser omentum. A lymph node obtained from the lesser curvature of the stomach contained caseating granulomas; a simultaneously obtained liver specimen was characterized by marked hepatocellular and canalicular cholestasis with mild ductopenia, but no fibrosis or cirrhosis. Staining of all tissues for acid-fast bacilli was negative. On the basis of the insidious nature of this patient's illness, the fever and lymphadenopathy, and the history of exposure to livestock, he was treated empirically for brucellosis with doxycycline and gentamicin; in addition, he was treated for H. pylori with amoxicillin, clarithromycin, and omeprazole. Despite the lack of a clear diagnosis, the patient was discharged to his home with a plan for close follow-up.
The normal bone marrow in the presence of circulating nucleated red cells and jaundice raises the possibility of a life-threatening disorder, such as sepsis. Brucellosis has protean manifestations, the most prominent of which is fever. Mild elevations in aminotransferases may occur, but the degree of hyperbilirubinemia seen in this case would be atypical. The caseating granulomas suggest the presence of tuberculosis, although caseation is not specific for mycobacterial infection; fungal and atypical bacterial infections should also be considered. Although it is not unreasonable to initiate empirical antibiotic therapy for suspected brucellosis, I am concerned about discharging the patient without a diagnosis, particularly since patients with hyperbilirubinemia and circulating nucleated red cells have a high in-hospital mortality rate.
At a follow-up visit one week later, the patient had no improvement in his symptoms. The white-cell count was 1100 per cubic millimeter, with an absolute neutrophil count of 200. The hemoglobin level was 11.5 g per deciliter; the platelet count was normal. Serologic tests for brucella and viral hepatitis were negative, as were tests for antinuclear, antineutrophil cytoplasmic, anti–smooth-muscle, and antimitochondrial antibodies. He was readmitted to the hospital. The fall in the white-cell count was attributed to the antibiotics, and it resolved after treatment with granulocyte colony-stimulating factor (G-CSF) and the discontinuation of his medications. He remained afebrile. Magnetic resonance cholangiopancreatography showed no dilatation of pancreatic or intrahepatic biliary ducts. Empirical, four-drug therapy for tuberculosis was begun, and the patient was discharged.
The neutropenia may have been a manifestation of worsening liver disease. Although the neutropenia resolved with the discontinuation of the antibiotics and the administration of G-CSF, the antibiotics that were given are not commonly associated with this adverse event. Of the four antibiotics, doxycycline and gentamicin are the least likely to be the offending agents. Brucellosis and tuberculosis are both associated with neutropenia. Although Brucella melitensis is the species that most commonly infects humans, infection with B. canis can also occur and does not result in antibodies against standard brucella antigens. Serum agglutination tests may also be negative early in the course of disease, necessitating repeated testing. Cultures for tuberculosis may be negative in the presence of miliary tuberculosis, but committing the patient to a six-month course of four-drug therapy for tuberculosis seems premature at this point. Brucella species are slow-growing organisms, and it may take as long as several weeks for the cultures to become positive. I would reinitiate treatment with doxycycline and gentamicin while awaiting the culture results.
The patient was readmitted two weeks later because of a white-cell count of 1700 per cubic millimeter and an absolute neutrophil count of 750. The level of alanine aminotransferase was 31 U per liter; that of aspartate aminotransferase, 41 U per liter; that of alkaline phosphatase, 350 U per liter; and that of total bilirubin, 11.9 mg per deciliter (203 μmol per liter). Repeated bone-marrow aspiration revealed no abnormalities. CT of the chest and abdomen showed enlarging retroperitoneal lymph nodes with new lymph nodes above the right main bronchus and in the left inguinal region (Figure 1). Surgical biopsy specimens of the hilar and inguinal lymph nodes revealed reactive lymphoid tissue and no granulomas; staining for acid-fast bacilli was negative. The neutropenia resolved after discontinuation of the antituberculosis treatment. The patient was discharged while taking empirical prednisone (60 mg daily orally) for presumed granulomatous hepatitis.
Figure 1. CT Scans of the Chest (Panel A) and Abdomen (Panel B).
Panel A shows a lymph node that is 2.5 cm in diameter and superior to the right main bronchus (arrow). Panel B shows a lymph node that is 2.4 cm in diameter and adjacent to the aorta and inferior vena cava (arrow).
The level of alkaline phosphatase has doubled since the initial laboratory evaluation, consistent with the presence of an infiltrative or cholestatic process. Granulomatous hepatitis is a syndrome characterized by prolonged recurrent fevers, myalgias, arthralgias, and granulomas in the liver and other organs. However, granulomatous hepatitis is a diagnosis of exclusion, and before empirical corticosteroid therapy is instituted, four to eight weeks of antituberculosis therapy is typically recommended. In the absence of a clear-cut diagnosis, I am concerned about the use of empirical therapy, particularly corticosteroids, given that infectious causes have not been ruled out. Instead, I would recommend repeating the liver biopsy, with cultures.
After vacationing in Mexico for a month, during which he continued corticosteroid therapy, the patient returned to the clinic with persistent weight loss and abdominal pain, but without fever. He was readmitted to the hospital and was found to have an alanine aminotransferase level of 818 U per liter, an aspartate aminotransferase level of 379 U per liter, an alkaline phosphatase level of 2132 U per liter, and a total bilirubin level of 36 mg per deciliter (616 μmol per liter). Endoscopic retrograde cholangiopancreatography showed relatively thin intrahepatic ducts, with a decreased number of ducts in the left hepatic lobe. He was transferred to a liver-transplantation center, where examination of a specimen from the second liver biopsy showed marked hepatocellular and canalicular cholestasis and no intact interlobular bile ducts.
This patient has a rapidly progressive intrahepatic cholestatic process. Drugs such as phenothiazines and environmental exposure to arsenic have been implicated in a syndrome characterized by progressive ductopenia: the vanishing bile-duct syndrome. However, on initial presentation the patient was not taking medications. Arsenic poisoning can result from exposure to contaminated well water or from the use of or exposure to arsenic-containing pesticides and herbicides and pressure-treated wood; as a gardener, he may have been exposed to these substances. Another herbicide, paraquat, has been associated with a cholangiodestructive process, although most cases of paraquat poisoning are the result of suicide attempts rather than accidental exposure. Lymphoma-associated ductopenia and idiopathic adulthood ductopenia have been described, including a variant of the latter that is characterized by symptoms of decompensated cirrhosis at presentation and a paucity of intrahepatic bile ducts on liver biopsy and that frequently requires orthotopic liver transplantation. The administration of corticosteroids might have been expected to improve a lymphoproliferative process, if only transiently. I admit that I remain puzzled. If I were caring for this patient, I would discontinue the prednisone and restart the antituberculosis therapy while contemplating what other types of tissue I should obtain to make a diagnosis.
After a progressive downhill course complicated by diffuse pulmonary infiltrates, clinical sepsis, and two cardiopulmonary arrests, the patient died. At autopsy, he was found to have had Hodgkin's disease in the mediastinal and mesenteric lymph nodes (Figure 2) and marked cholestasis and ductopenia in the liver (Figure 3). A suppurative, necrotizing fungal pneumonia was present throughout the lungs. There was no identifiable pulmonary malignant tumor or evidence of hepatic fungal infection. A diagnosis of vanishing bile-duct syndrome due to Hodgkin's disease was made. A retrospective review of all previous biopsy specimens showed no evidence of Hodgkin's disease.
Figure 2. Specimen of an Abdominal Lymph Node Obtained at Autopsy, Showing Infiltrates of Lymphocytes, Plasma Cells, Eosinophils, and Reed–Sternberg Cells (Arrow).
Figure 3. Biopsy Specimen of the Liver Obtained from the Patient at Autopsy (Panel A) and Biopsy Specimen of Normal Liver (Panel B) (Hematoxylin and Eosin).
Bile ducts are absent in the portal tract shown in Panel A, whereas a bile duct is present in the portal tract shown in Panel B (arrow).
Cholestasis can occur in patients with Hodgkin's disease because of extrahepatic biliary obstruction from extrinsic compression by enlarged lymph nodes, direct hepatic infiltration, or a rare but well-recognized paraneoplastic process that is characterized by intrahepatic cholestasis, with or without bile-duct paucity. In retrospect, what is striking about this case is the lack of evidence of Hodgkin's disease on the antemortem lymph-node examination. Clearly, a diagnosis of Hodgkin's disease should be considered in any patient with unexplained intrahepatic cholestasis, particularly a febrile cholestatic illness. Perhaps this case best illustrates the problem of relying too much on too little information with which to base empirical therapy.
Commentary
Vanishing bile-duct syndrome is a rare entity characterized by the progressive loss of small intrahepatic bile ducts, leading to cholestatic hepatitis and often death. Ductopenia may be due to immunologic, infective, vascular, drug-induced, paraneoplastic, or idiopathic causes.1 Since cholangiography typically shows no abnormalities and the histologic loss of bile ducts may be subtle, the lesion is often missed on initial pathological review.1
The association of vanishing bile-duct syndrome with Hodgkin's disease has been well established and is most likely a paraneoplastic phenomenon.1,2,3,4,5 In this syndrome, there is no evidence of hemolysis, extrinsic compression of the extrahepatic ducts, or tumor infiltration on histologic examination of the liver. Unlike other causes of vanishing bile-duct syndrome, those that occur in association with Hodgkin's disease may be reversible if chemotherapy induces a remission of the cancer.2 Cholestatic hepatitis has been described in several case reports as appearing several weeks before the Hodgkin's disease became clinically apparent.2,6 Although Hodgkin's disease reportedly has remained undiagnosed after an initial bone marrow biopsy,2 we are not aware of reports of a similar difficulty in detecting Hodgkin's disease after numerous lymph-node and bone marrow biopsies, as in this case. This delay in diagnosis, compounded by the subtlety of the bile-duct loss on the initial liver-biopsy specimen, resulted in a substantial delay in identifying the cause of the patient's jaundice.
When a patient's condition worsens despite the use of empirical therapy, clinicians must decide whether to continue the empirical therapy, change the empirical therapy, or repeat diagnostic testing. Such decisions, which must often be made in practice, ultimately require clinical judgment to determine whether the presumed diagnosis is correct and to evaluate the risks of treating or not treating the patient. In this case, the discussant was concerned about the empirical use of corticosteroids. The use of that therapy, together with certain other factors (including the patient's unrecognized underlying lymphoma, the frequent hospitalizations, and the use of intravenous catheters and antibiotics), probably contributed to the ultimate development of fungal pneumonia.
Because a mathematical decision-tree analysis is too complex for physicians' daily use in establishing the best course of action, they tend to make decisions in an incremental or consecutive fashion.7 This approach allows them to use whatever limited data are available, including personal experience and the patient's response to therapy, to decide the next treatment. In this way, each course of failed empirical therapy serves as a new piece of diagnostic information and may help exclude another disorder from the differential diagnosis.
However, when using a consecutive approach to problem solving, clinicians may fail to reanalyze the data on which previous decisions were made. In this case, the lack of an identifiable malignant tumor in numerous biopsy specimens gave the clinicians the false sense that cancer had definitively been ruled out. Thus, infectious and inflammatory conditions became the two most likely categories of disease, leading to a series of empirical treatments aimed at the most probable cause. Ideally, clinicians should retrospectively evaluate the clinical data on which they have based their previous decisions and consider repeating diagnostic studies if the trial-and-error approach seems to be failing.
In this case, the histologic identification of vanishing bile-duct syndrome and the recognition of the potentially latent appearance of Hodgkin's disease in association with this syndrome might have led to earlier diagnosis and potentially lifesaving treatment. This case reminds us of the link between the vanishing bile-duct syndrome and Hodgkin's disease and the frustration and consequences that result from being empirically incorrect.
Supported by a Career Development Award from the Health Services Research and Development Program of the Department of Veterans Affairs (to Dr. Saint) and a Patient Safety Developmental Center Grant (P20-HS11540, to Dr. Saint) from the Agency for Healthcare Research and Quality.
No potential conflict of interest relevant to this article was reported.
We are indebted to Donald Houghton, M.D., Sandra White, M.D., and Tracy Funk of Oregon Health and Sciences University Department of Pathology for reviewing all the biopsy and autopsy specimens pertaining to this case.
Source Information
From the Department of Medicine, Legacy Emanuel and Legacy Good Samaritan Hospitals, Portland, Oreg. (A.S., D.J.G.); and the Department of Medicine, Ann Arbor Veterans Affairs Medical Center (S.S., R.H.M.); Patient Safety Enhancement Program, Ann Arbor Veterans Affairs Medical Center and University of Michigan Health System (S.S.); and the Department of Internal Medicine, University of Michigan Medical School (S.S., R.H.M.) — all in Ann Arbor.
Address reprint requests to Dr. Schmitt at the Legacy Emanuel Hospital and Health Center, 2801 N. Gantenbein Avenue, Suite 4100, Portland, OR 97227, or at aschmitt@lhs.org.
References
Hubscher SG, Lumley MA, Elias E. Vanishing bile duct syndrome: a possible mechanism for intrahepatic cholestasis in Hodgkin's lymphoma. Hepatology 1993;17:70-77.
Crosbie OM, Crown JP, Nolan NPM, Murray R, Hegarty JE. Resolution of paraneoplastic bile duct paucity following successful treatment of Hodgkin's disease. Hepatology 1997;26:5-8.
Birrer MJ, Young RC. Differential diagnosis of jaundice in lymphoma patients. Semin Liver Dis 1987;7:269-277.
Warner AS, Whitcomb FF. Extrahepatic Hodgkin's disease and cholestasis. Am J Gastroenterol 1994;89:940-941.
Jansen PL, van der Lelie H. Intrahepatic cholestasis and biliary cirrhosis associated with Hodgkin's disease. Neth J Med 1994;44:99-102.
Lieberman DA. Intrahepatic cholestasis due to Hodgkin's disease: an elusive diagnosis. J Clin Gastroenterol 1986;8:304-307.
Moskowitz AJ, Kuipers BJ, Kassirer JP. Dealing with uncertainty, risks, and tradeoffs in clinical decisions: a cognitive science approach. Ann Intern Med 1988;108:435-449.(Amy Schmitt, M.D., Daniel)