hemostasis). This alteration in my approach to such lesionshas already yielded clinical benefit in a patient whose non-bleeding lesion was readily identified at subsequent surgery.Although long-term hemostasis can now be achieved endo-scopically in the majority of bleeding Dieulafoy-like le-sions, a present challenge is to optimize care and improveoutcome in patients in whom this first-line therapy fails.Routinely “marking the spot” may be a step in this direction,and will undoubtedly assist our gastrointestinal endoscopicand surgical colleagues, when their expertise is needed inmanaging these often frustrating cases of recurrent gastro-intestinal hemorrhage.

Joseph C. Yarze, M.D., F.A.C.P., F.A.C.G.Gastroenterology Associates of Northern New York

Glens Falls, New York

Reprint requests and correspondence:Joseph C. Yarze, M.D.,F.A.C.P., F.A.C.G., Gastroenterology Associates of Northern NewYork, P.C., Five Irongate Center, Glens Falls, NY 12801.

Received Aug. 31, 2000; accepted Sep. 13, 2000.

Acute Liver Failure DuringLamivudine Treatment in a HepatitisB Cirrhotic PatientTO THE EDITOR: Lamivudine has been shown to beeffective in the treatment of chronic hepatitis B. Generallylamivudine is well tolerated and has a good safety profile(1). We report a case of reactivation of chronic hepatitis Bin a patient treated long-term with lamivudine.

A 47-yr-old white man with chronic hepatitis B (HBV)infection (biopsy stage: cirrhosis with associated severechronic hepatitis; HBcAg in 90% of hepatocytes; HBeAgnegative; HBV-DNA 300 pg/ml; ALT 5–7 times abovenormal limit) who received a previous standard course ofrecombinant interferon-a therapy in 1997, was started onlamivudine 100 mg/day monotherapy in January 1998. De-spite lamivudine treatment, ALT and HBV-DNA levelswere persistently elevated, without signs of side effects. Thepatient stopped lamivudine in March 1999 because of thelack of clinical response. Four weeks after discontinuationof lamivudine, ALT and HBV-DNA levels rose significantlyto 10–15 times above the normal limit.

The patient was again given lamivudine treatment. Oneweek later, ALT increased to 1658 U/L, bilirubin to 1028mg/dl, and INR to 4.

We excluded coinfection with hepatitis A virus, hepatitisC virus, and superinfection with hepatitis D virus, as well asautoimmune hepatitis. Clinical status rapidly deterioratedwith fever, ascites, and signs of hepatic encephalopathy. Thepatient was treated with supportive therapy and prednisone,assuming a severe immunological reaction to HBV-infectedhepatocytes, as lamivudine can restore T cell response in

chronic hepatitis B (2). Despite treatment the patient dete-riorated dramatically and died 5 days after clinical onset.

Cases of spontaneous reactivation of HBV infection dur-ing lamivudine treatment have been described regardingonly naive patients, but a relationship with lamivudine hasnot been proved (3). Some cases have responded well toimmunosuppressive therapy, suggesting an immunologicalreaction against HBV-infected hepatocytes. Our case is thefirst, to our knowledge, regarding a patient with advancedliver disease. This is important to note because of severalupcoming trials employing lamivudine in cirrhotic patients.

Although this case cannot demonstrate that Lamivudinein HBV cirrhotic patients may cause acute liver failurebecause of a massive reinfection of hepatocytes, we suggestcareful monitoring of biochemical parameters and earlyinitiation of immunosuppressive therapy.

Raffaele Bruno, M.D.Paolo Sacchi, M.D.Carlo Filice, M.D.

Gaetano Filice, M.D.Division of Infectious and Tropical Diseases

IRCCS S. Matteo HospitalUniversity of Pavia

Pavia, Italy

REFERENCES

1. Dienstag JL, Shiff ER, Wright TR, et al. Lamivudine as initialtreatment for chronic hepatitis B in the United States. N EnglJ Med 1999;341:1256–63.

2. Boni C, Bertoletti A, Penna A, et al. Lamivudine treatment canrestore responsiveness in chronic hepatitis B. J Clin Invest1998;102:968–75.

3. Honkoop P, de Man RA, Heijtink RA, et al. Hepatitis B reac-tivation after lamivudine. Lancet 1995;346:1156–7.

Reprint requests and correspondence:Raffaele Bruno, Divi-sione Malattie Infettive e Tropicali, IRCCS Policlinico S. Matteo,Via Taramelli, 5, 27100 Pavia, Italy.

Received Sep. 7, 2000; accepted Sep. 13, 2000.

Mesenteric Arteriovenous FistulaCausing Jejunal Varices andRecurrent GI BleedingTO THE EDITOR: Mesenteric arteriovenous fistulas arerare. They usually complicate penetrating abdominal traumaor gastrointestinal surgery. We report a patient who pre-sented with recurrent severe, obscure gastrointestinal bleed-ing 4 yr after a small bowel resection for a gunshot wound.He was found to have a mesenteric arteriovenous fistulaproducing large jejunal varices. Recurrent severe bleeding

265AJG – January, 2001 Letters to the Editor

ceased after surgical resection of the fistula and the involvedjejunal segment.

A 21-yr-old man presented for evaluation of painlessmelena. He was well until 4 yr earlier when he sustainedgunshot wounds to his abdomen, requiring a segmentaljejunal resection. Three years later he presented elsewherewith painless melena requiring transfusions. No bleedingsource was found. Physical examination was significant forpallor and an abdominal bruit. There was no organomegaly,signs of chronic liver disease, or hyperdynamic circulation.Stools were melenatous and strongly positive for occultblood. Hemoglobin was 5.8 g/dl. Iron indices were consis-tent with iron deficiency. Liver biochemistries and coagu-lation profile were normal. Upper endoscopy with pushenteroscopy showed onlyHelicobacter pylori–positive gas-tritis. Colonoscopy and ileoscopy were normal. Small bowelfollow-through x-ray was normal and a Meckel’s scan wasnegative. The patient was transfused 6 U of packed redblood cells. He was treated forHelicobacter pylori andgiven iron orally. He then presented again 6 months laterwith another episode of weakness and painless melena.Hemoglobin was 8.7 g/dl. Abdominal angiogram (Fig. 1A,B) revealed a mesenteric arteriovenous fistula with fillingfrom the jejunal artery feeding into a large grape-like clusterof mesenteric varices.

At surgery, a palpable thrill was noted at the fistula site.Ligation of the feeding vessel was performed, with visiblecollapse of the mesenteric vessels. However, the patientpresented again 6 months after with severe hemorrhage anda hemoglobin of 6 g/dl. Surgical resection of the involvedsmall bowel (57 cm) and mesenteric vessels was then per-formed. No recurrence of bleeding has occurred with fol-low-up of 2 yr.

Our patient developed a mesenteric arteriovenous fistulainvolving branches of the jejunal artery and vein after pen-etrating abdominal trauma (gunshot wounds) and jejunalresection. Pathogenesis requires simultaneous injury and/orlocal infection involving both the mesenteric artery andvein, with resultant fistula formation. Our patient manifestedwith recurrent obscure gastrointestinal bleeding from seg-mental portal hypertension: congestive enteropathy (conges-tive jejunopathy) and/or intraluminal rupture of segmentalsmall bowel varices. He did not manifest symptoms, anysigns of a more generalized portal hypertension, hyperdy-namic circulation, or mesenteric steal syndrome. Because ofmultiple collateral arteries and failure of surgical ligation ofthe main feeding artery to control the bleeding, surgicalresection was required as definitive treatment.

Arterioportal fistulas may involve intrahepatic or extra-hepatic vessels. In a recent review of the literature onarterioportal fistulas, 15% were congenital and 85% wereacquired (1). Congenital causes include arteriovenous mal-formations, Osler-Weber-Rendu syndrome, Ehlers-Danlossyndrome, and aneurysms (1). Acquired causes include

trauma, iatrogenic procedures, tumor, aneurysms, and idio-pathic (1). Gastrointestinal bleeding occurs in 33%, ascitesin 26%, high output heart failure in 5%, and symptoms ofmesenteric steal syndrome in 5% of patients (1). Intrahe-patic arterioportal fistulas can lead to gastrointestinal bleed-ing via hemobilia or as a result of generalized portal hyper-tension (1–4). Extrahepatic arterioportal fistulas can lead togastrointestinal bleeding via segmental or generalized portalhypertension, erosion of the fistula into the lumen of thebowel, ulceration of overlying congested mucosa, and isch-emic bowel from mesenteric steal syndrome (1–5).

Treatment of symptomatic arterioportal fistulas requires a

Figure 1. Angiogram showing mesenteric arteriovenous fistulacausing massive segmental mesenteric varices. Note dilated mes-enteric arteries (A, arterial phase) and veins (B), venous phase).

266 Letters to the Editor AJG – Vol. 96, No. 1, 2001

case-by-case decision and can be managed by radiologicalintervention, surgical treatment, or both (1–5).

Harish Patil, M.D.Keith Weiler, M.D.

Michael Jachec, M.D.William Hopkins, M.D.Virendra Mathur, M.D.

Charles Berkelhammer, M.D.Departments of Medicine, Radiology, and Surgery

Christ Hospital and Medical CenterUniversity of Illinois

Oak Lawn, Illinois

REFERENCES

1. Vauthey JN, Tomczak RJ, Helmberger T, et al. The arteriopor-tal fistula syndrome: Cliniocopathologic features, diagnosis andtherapy. Gastroenterology 1997;113:1390–401.

2. Pietri J, Remond A, Reix T, et al. Arterioportal fistulas: Twelvecases. Ann Vasc Surg 1990;4:533–9.

3. Strodel WE, Eckhauser FE, Lemmer JH, et al. Presentation andperioperative management of arterioportal fistulas. Arch Surg1987;122:563–71.

4. Alkin C, Sahin T, Oguz P, et al. A case report of congenitalintrahepatic arterioportal fistula. Am J Gastroenterol 1999;94:523–5.

5. Capron JP, Gineston JL, Remond A, et al. Inferior mesentericarteriovenous fistula associated with portal hypertension andacute ischemic colitis. Successful occlusion by intra arterialembolization with steel coils. Gastroenterology 1984;86:351–5.

Reprint requests and correspondence:Charles Berkelhammer,M.D., 9921 Southwest Highway, Oak Lawn, IL 60453.

Received Aug. 22, 2000; accepted Sep. 5, 2000.

Percutaneous Endoscopic Gastrostomyin Patients With DementiaTO THE EDITOR: We read with great interest the article bySanderset al. in your recent issue (1). We totally agree withtheir conclusion advising against gastrostomy feedings inselected patients with dementia because of the high mortal-ity rate. We would like to point out a population-basedcohort study of 97 patients referred for percutaneous endo-scopic gastrostomy (PEG) published by us earlier from ourinstitution (2). Based on the results in this study and asubsequent study involving surgical gastrostomy at our in-stitution (3), we would like to bring to the attention of thereaders the following points.

First, the 30-day mortality (22%) and 1.5-yr mortality(65%) in our study are similar to the 30-day (28%) and 1-yr(63%) mortality reported by Sanderset al. However, the30-day and 1-yr mortality rates in their dementia group aremuch higher (54% and 90%, respectively). In contrast, thesurvival in our patients with diseases of the central nervoussystem (excluding vascular diseases) was not significantlyworse than in patients having tubes placed for other indica-

tions for PEG (survival adjusted for age, gender, comorbidconditions, level of consciousness, and activity level).

Second, the major complication rate was 2.5% and pneu-monia was attributed as the cause of death in 27% of thepatients in the study by Sanderset al. We had a complicationrate of 70%, but 88% of them were minor. The major causesof death were pneumonia, heart disease, and cardiovasculardisease.

Third, although the Cox proportional hazards analysisindicated that age and group (according to diagnosis) wereindependent predictors of mortality in the study by Sanderset al., we noted the mean age of the dementia group in thatstudy was the highest among four groups at 77.1 yr. It wasnot clear whether the “age effect” was similar among allgroups, in particular for the dementia group. We observed inour study that older age and hypoxemia were significantlyassociated with survival (but not, for example, indication forPEG or level of activity) based on a similar proportionalhazards regression analysis. It would have been helpful toassess the impact of age on survival if Sanderset al. hadprovided Kaplan-Meier estimates of survival by age group(e.g., 40–59 yr, 60–79 yr,$80 yr) separately for each oftheir four major groups. If age had no differential impact onsurvival in the dementia group, the survival curve in Figure1 of their article would presumably show similar patterns inthe four major groups.

The observations and recommendations made by Sanderset al. are very pertinent, and merit serious consideration byall health care personnel involved in decision making forpatients with dementia.

L. R. Bergstrom, M.D.V. Santhi Swaroop, M.D.

General Internal MedicineMayo Clinic

Rochester, Minnesota

A. R. Zinsmeister, Ph.D.Section of Biostatistics

Department of Health Sciences ResearchMayo Clinic

Rochester, Minnesota

REFERENCES

1. Sanders DS, Carter MJ, D’Silva J, et al. Survival analysis inpercutaneous endoscopic gastrostomy feeding: A worse outcomein patients with dementia. Am J Gastroenterol 2000;95:1472–5.

2. Taylor CA, Larson DE, Ballard DJ, et al. Mayo Clin Proc1992;67:1042–9.

3. Bergstrom LR, Larson DE, Zinsmeister AR, et al. Utilizationand outcomes of surgical gastrostomies and jejunostomies in anera of percutaneous endoscopic gastrostomy: A population-based study. Mayo Clin Proc 1995;70:829–36.

Reprint requests and correspondence:L. R. Bergstrom, M.D.,Area General Internal Medicine, Mayo Clinic, Rochester, MN55905.

Received Aug. 17, 2000; accepted Aug. 28, 2000.

267AJG – January, 2001 Letters to the Editor