negative histology for WD, were complaining of dyspepsia (173) with the remainder (37) having malabsorption. A second, smaller group of 132 patients, again with negative histology but strong clinical suspicion for WD, had small intestinal biopsies submitted for PCR analysis. In a vindi- cation for the academic pathologist, there was complete concordance between negative histology and negative PCR in all the specimens. The authors conclude that there is no reason for PCR-based T. whipplei testing in patients with negative histology, even for those in whom the clinical suspicion is high. What lessons can we learn from this study? Firstly, WD is extremely rare and is diagnosed infrequently, even in patients with malabsorption evaluated in specialized gastro- enterology clinics. Secondly, PCR-based T. whipplei test- ing, as done by this group, is not associated with false positive results, thereby dispelling the notion that T. whip- plei might be a commensal in the intestine. Lastly, WD might not involve the bowel in up to 15% of cases (so called dry forms of the disease) (2), and testing of a symptomatic extraintestinal site by PCR is indicated. Citing previous work from their institution, these authors describe patients with negative intestinal biopsies whose diagnosis of WD was made by PCR of other infected tissues. This included CSF, a lymph node, ocular vitreous fluid, and a heart valve. Most practicing physicians will never diagnose WD in their working lifetime, so it might seem such research will not repay the effort invested in it. However, it should be remembered that WD is uniformly fatal if left untreated and is a diagnosis never to be missed. Mark Flemmer, M.D. Terri Esterowitz, M.D. Eastern Virginia School of Medicine Norfolk, Virginia REFERENCES 1. Relman DA, Schmidt TM, MacDermott RP, et al. Identification of the uncultured bacillus of Whipple’s disease. N Engl J Med 1992;327:293–301. 2. Misbah SA, Ozols B, Franks A, et al. Whipple’s disease without malabsorption: New atypical features. J Med 1997;90:765–72. Surveillance for Hepatocellular Carcinoma: Does It Work? Bolondi L, Sofia S, Siringo S, et al. Surveillance programme of cirrhotic patients for early diagnosis and treatment of hepatocellular carcinoma: A cost effectiveness analysis. Gut 2001;48:251–9 Bolondi et al. evaluated the cost effectiveness of surveil- lance of hepatocellular carcinoma (HCC). A cohort of 313 Italian patients with cirrhosis was enrolled in an HCC sur- veillance program that consisted of ultrasonography (US) and serum -fetoprotein (AFP) performed every 6 mo. Hep- atitis C was present in 64%, hepatitis B (17%), and alcohol abuse (25%). This cohort was compared to a control group of 104 consecutive patients referred from other facilities for incidentally detected HCC. The two main outcomes exam- ined were eligibility of patients for treatment as well as survival. Additional outcomes were the overall cost of sur- veillance program; cost per treatable HCC; and cost per year of life saved. During a follow-up period averaging 56 months, 61 patients (20%) in the surveillance program de- veloped HCC (incidence 4.1% year). Only 69% of patients with HCC were treated as follows: 9% had surgical resec- tion, 26% liver transplantation, 24% percutaneous ethanol injection, and 43% had transarterial chemoembolization [TACE]. The cumulative survival rate of the patients with HCC detected in the surveillance program was significantly longer than that of the controls (p = 0.02). The median survival for patients with HCC in the surveillance and control groups was 30 and 15 months, respectively. Multi- variate analysis showed surveillance to be an independent predictor of longer survival. Elevated AFP, Child-Pugh classes B and C, and male sex were found to be independent risk factors for developing HCC. The overall cost of the surveillance program was $753,226. The cost per treatable HCC was $17,934, and the cost for year of life saved was $112,993. The cost authors concluded that their HCC sur- veillance program in patients with cirrhosis consumed a large number of resources and offered little benefit in patient survival. (Am J Gastroenterol 2002;97:2676 –2677. © 2002 by Am. Coll. of Gastroenterology) COMMENT The incidence of HCC has been rising in the United States and the survival of patients with HCC continues to be dismal (1). Most cases of HCC arise in patients with cirrhosis; this theo- retically allows for the opportunity to apply surveillance tests to detect HCC at an early and thus potentially curable stage. Tests such as AFP and US have low diagnostic accuracy when used in HCC surveillance. The receiver operating characteristics of AFP as a surveillance test are such that lowering the cutoff value of AFP for HCC diagnosis in- creases its sensitivity but lowers the specificity. A French trial that examined 118 patients with Child-Pugh A or B cirrhosis found that changing the AFP cutoff value from  15 ng/mL to  100 ng/mL increased the sensitivity of AFP from 21% to 50% but decreases the specificity from 93% to 86% (2). Similar observations were made in a prospective study from Japan in which 260 patients with cirrhosis were screened for HCC; increasing the cutoff value for AFP from 20 ng/mL to 100 ng/mL increased the sensitivity from 13% to 39% and decreased the specificity for detecting HCC from 97% to 76% (4). Ultrasound is superior to AFP for the early detection of HCC (3, 4). For example, ultrasound demonstrated higher diagnostic accuracy (AFP was below diagnostic levels in all patients who had HCC by US) than 2676 World Literature Review AJG – Vol. 97, No. 10, 2002