Purpose: Patients who are taking chronic glucocorticoids (GC) for avariety of disease conditions can lose significant amounts of bone massvery quickly. It is estimated that this bone loss leads to fractures in 30–50%of patients. Therapies that prevent bone loss or increase bone mass mayprevent secondary osteoporosis from developing in patients on GC therapy.We have conducted a study to determine the effect of risedronate onfracture rates and spinal BMDin those patients on glucocorticoid therapy.Methods: Men and women, 18–85 years of age (n � 508) receivingglucocorticoids (�7.5 mg/day) were enrolled in two clinical studies withsimilar protocols. The prevention study enrolled patients who had been onGC therapy for 3 months or less. The treatment study enrolled patients whohad been on GC therapy for more than 6 months. All patients receivedplacebo or risedronate 2.5 or 5 mg daily for 1 year while continuing GCtherapy. Patients in the prevention study received 500 mg/day of elementalcalcium while those in the treatment study received 1000 mg of elementalcalcium and 400 IU vitamin D daily. Change in lumbar spine BMD frombaseline to 1 year was the primary efficacy endpoint.Results: At 12 months, sixteen percent of patients had new vertebralfractures in the placebo groups, compared to 7% and 5% in the 2.5 and 5mg/day risedronate-treated groups respectively. A statistically significant70% vertebral fracture risk reduction vs. placebo was observed (p � 0.01)when risedronate 5 mg/day data from the two studies were pooled. Asignificant increase in BMD at the lumbar spine (2.9%, p � 0.001), femoralneck (2.8%, p � 0.001) and femoral trochanter (2.8%, p � 0.001) com-pared to placebo was observed in the risedronate-treated (5 mg/day) groups.Adverse events were similar between placebo and risedronate-treatedgroups. Upper GI adverse events were also similar between the placebo andrisedronate-treated groups. There was a slightly higher incidence of backpain and arthralgia in risedronate-treated patients, but this did not lead tomore frequent patient discontinuation.Conclusions: Risedronate is an effective and well-tolerated therapy forreducing vertebral fracture risk in patients either initiating or on mainte-nance GC therapy.

867

Role of TGF-�1 and TGF-� type II receptor in gastric cancerJu Sung Kim, MD 1, Hee Jung Son, MD1*, Dong Il Park, MD1, SangYong Song, MD2, Won Hyeok Choe, MD1, Yun Jeong Lim, MD1, Sang-Jong Park, MD1, Jae J. Kim, MD1, Young-Ho Kim, MD1, Poong-LyulRhee, MD1, Seung Woon Paik, MD1, Jong Chul Rhee, MD1 and KyooWan Choi, MD1. 1Medicine, Samsung Medical Center, SungkyunkwanUniversity School of Medicine, Seoul, Korea; and 2Pathology, SamsungMedical Center, Sungkyunkwan University School of Medicine, Seoul,Korea.

Purpose: Transforming growth factor-beta is a potent inhibitor of epithelialcell growth. However, carcinoma cells, unlike normal cells, can escapefrom negative regulation by TGF-beta through lack of expression or mu-tation of TGF-beta receptor gene. In this study, we investigated the role ofTGF-beta 1 and TGF-beta type II receptors in the progression of gastriccancer.Methods: We analyzed TGF-beta 1 and TGF-beta type II receptor mRNAexpression semi-quantitatively, measured by comparative RT-PCR usingGAPDH, in 23 patients who underwent gastric resection for gastric cancer.We analyzed the relationship between the clinicopathologic findings andthe level of the TGF-beta 1 and TGF-beta type II receptor mRNA expres-sion in carcinoma tissues and in adjacent normal tissues of gastric cancer.Results: TGF-beta 1 and TGF-beta type II receptor mRNA were expressedin all of the carcinoma tissues and adjacent normal tissues without statis-tical difference in the level of the expression. The level of TGF-beta 1mRNA expression was higher in patients with early gastric cancer, negativelymph nodes or negative perineural invasion. There was no significantcorrelation between the level of TGF-beta 1 mRNA expression and severalparameters such as age, gender, tumor size, differentiation, Lauren’s clas-sification, and vascular invasion, there was no significant correlation be-tween the level of TGF-beta type II receptor mRNA expression and several

prognostic variables described above. There was significant correlationbetween the level of TGF-beta 1 and TGF-beta type II receptor mRNA incarcinoma tissues.Conclusions: The above data indicates that TGF-beta 1 may contribute inthe early stage of gastric carcinogenesis. Further studies are required toclarify the role of TGF-beta 1 in gastric carcinogenesis.

868

Predictive factors of short-term mortality after PEG placementNejat Kiyici, MD, Yusuf Z Tatli, MD, C. S. Pitchumoni, MD, FACG*,Hilary Hertan, MD, FACG and Edward P Norkus, Ph.D.1Gastroenterology, Our Lady of Mercy Medical Center, Bronx, NewYork, United States.

Purpose: To identify the predictive factors for short-term mortality afterPEG placement using Charlson co-morbidity index.Methods: In this prospective cohort study short-term mortality was as-sessed in 57 patients referred for PEG (percutaneous endoscopic gastros-tomy) placement. Patient demographics, admitting diagnosis, indication forPEG, DNR status and comorbid conditions were analyzed. 19 differentvariables were used to calculate Charlson co-morbidity index (CCI).Nu-tritional status was evaluated by serum albumin, total lymphocyte count,total cholesterol and weight. A standard rehabilitation medicine scale wasused to evaluate functional status of all patients. The mean pre-PEGalbumin level was 2.6�/� 0.6 (SD)g/dL and mean age was 80 �/�12(SD). Patients were followed for 6 months.Results: Dementia, CVA and ventilatory assistance were the most commonindications for PEG placement. Overall mortality at 30 days was 14 %,reaching 40 % at 6 months. Logistic regression analysis was used toexamine a theoretical model for survival at 1 and 6 months. We found thatpre-PEG albumin level less than 2.5 g/dL, COPD and a CCI more than 3predicted poor survival 6 months after PEG placement. These analysisdetermined that the likelihood of death increased by 8 % (P � 0.004) withevery 0.1 g/dL decrease in pre-PEG serum albumin less than 2.5 g/dL, and13.9 fold increased likelihood of death (P � 0.009) if the CCI was greaterthan 3 at the time of PEG. This model also examined survival at 1 monthafter PEG and found that only pre-PEG albumin level less than 2.5 g/dLwas significant predictor of poor survival. Additional independent variables(Age, sex, DNR status, functional status, pre-PEG hemoglobin, total lym-phocyte count and weight) were examined in a step-wise fashion to theregression model. None were found to have any predictive effect onsurvival.Conclusions: Serum albumin level of less than 2.5 g/dL, COPD and CCImore than 3 appear to be sensitive predictors for poor survival after PEGplacement. Our findings suggest that these predictors may be appropriateidentifiers of patients with low probability of survival following PEGplacement.

Independent variables Odds-Ratio 95% confidence interval P-value

Age 1.0 0.91–1.0 0.143Sex 0.7 0.1–4.6 0.769Pre-PEG albumin 0.08 0.01–0.45 0.004CCI 1.7 1.1–2.5 0.009COPD 26.6 1.3–549.0 0.034

869

Diagnosis of irritable bowel syndrome in managed care: missedopportunity?Kevin Knight, MD1, Keiko Higuchi, MPH1, Alisa Wilson, Ph.D.1, JohnWong, B.S.1, Chiun-Fang Chiou, Ph.D.1, Victoria Barghout, M.S.P.H.2,Feride Frech, MPH2, George Longstreth, MD3 and Joshua Ofman,MD1*. 1Pharma and Research, Zynx Health Inc., Beverly Hills, CA,United States; 2Health Economics and Outcomes Research, NovartisPharmaceutical Corporation, East Hanover, NJ, United States; and3Southern California Kaiser Permanente Medical Group, San Diego,CA, United States.

S272 Abstracts AJG – Vol. 96, No. 9, Suppl., 2001

Purpose: This analysis estimates the likelihood of identifying irritablebowel syndrome (IBS) patients with administrative encounter data and asurvey using the Rome I symptom criteria in a sample of patients havinghad a flexibible sigmoidoscopy for any reason.Methods: A survey including the ROME I symptom criteria for IBS wassent to 6,500 enrollees in a large managed care organization (MCO) whohad received a flexible sigmoidoscopy during the year 2000. Patients wereidentified as IBS-positive or negative based on Rome I criteria. Encounterdata for the prior two-year period were examined. Patients without prac-titioner visits or hospitalizations were eliminated. Patients were classifiedas having definite IBS (at least one administrative code for IBS); possibleIBS (at least one administrative code for related GI conditions such asabdominal pain, constipation, diarrhea, dyspepsia, and reflux); or no IBS.Results: Survey responses were received from 2,613 patients (40.2%response rate), of whom 2,594 had at least one practitioner visit or hospi-talization. 619 patients (23.9%) were IBS positive and 117 (4.5%) had atleast one claim for IBS. Among IBS positive respondents, 70 had a definiteIBS code (sensitivity � 11.3%), while 191 (30.9%) had codes indicatingpossible IBS. Among IBS negative respondents, 47 of 1,975 had codesindicating definite IBS and 335 (17.0%) had codes indicating possible IBS(specificity � 97.6%). Even among IBS positive respondents consideredhigh utilizers (10� visits/hospitalizations (n � 458)), only 58 (12.7%) hada code indicating definite IBS, while 159 (34.7%) had a code indicatingpossible IBS.

IBS DiagnosisBased on Rome I

IBS Diagnosis Based on Encounter Data

Definite IBS(N � 117)

Possible IBS(N � 526)

No IBS(N � 1,951)

Positive (N � 619) 70 (11.3%) 191 (30.9%) 358 (57.8%)Negative (N � 1,975) 47 (2.4%) 335 (17.0%) 1,593 (80.6%)

Conclusions: Among a sample of MCO patients having undergone flexiblesigmoidoscopy for any reason, administrative encounter data identified asmall proportion of those with IBS, even among high utilizers of health careservices. Even when other GI conditions that could be related to IBS areincluded, fewer than half of patients who meet ROME I criteria for IBSwould have been identified using administrative data. These findings sug-gest that IBS is either substantially under-recognized and/or that appropri-ate diagnostic codes are often not applied to IBS patients. This research wassupported by a Research Grant from Novartis.

870

Management of chronic constipation in a community hospitalHolger Kranich, Affan Quadri, Geetha Ganesan, Affi Aboud andThomas Puetz*. 1Gastroenterology, University Of WI MilwaukeeCampus SSMC, Milwaukee, WI, United States.

Purpose: Patients with chronic constipation make frequent visits to Emer-gency department (ED) and Primary care providers (PCPs). This may bedue to the lack of proper long term care plans. Aims: To look at thediagnostic work-up and treatment offered to patients with chronic consti-pation who make frequent visits to the ED or PCPs.Methods: All patients with chronic constipation making 3 or more visits toED or PCPs, in a Community Hospital were identified. A structuredquestionnaire was used to record the demographics, coexisting conditions/medications affecting constipation, diagnostic work-up and treatment of-fered to these patients for each visit. In addition any long term care plan andpatient’s compliance was also recorded.Results: Forty-three patients had 210 visits to ED or PCPs between 1995and 1999 for chronic constipation. There were 26 females and 17 maleswith mean age of 61. Most common presentation was abdominal pain. Thediagnostic evaluation included Abdominal X-rays 89(84%), Out patientColonoscopy/Flexible sigmoidoscopy in an open access system 16(15%)and SITZMARKS study 1 (0.9%). Medications used, which may contributeto constipation, were Calcium Channel blockers, Antipsychotics, Antide-

pressants, and Narcotic pain medications. Most of the patients were con-tinued on these medications. Prophylactic treatment offered was Osmoticlaxatives 20%, Stool softeners 14%, Stimulant laxatives 12%, Fiber 10%,Enemas/Laxative suppositories 9 %, and Golyte/Miralax 2%. Thirty-threepercent of the patients were not given any prophylactic treatment. Only11.4 % patients were compliant to the treatment given on previous visits.Conclusions: Patients with chronic constipation make frequent visits toEmergency departments and Primary Care Providers. Better characteriza-tion of patients with proper diagnostic work-up, long term care plans andimprovement of noncompliance to treatment would be needed to over comethis problem.

871

Evidence-based analysis of the benefit of esomeprazole in thetreatment of erosive esophagitis (EE)Jeffrey G Levine*, Clara Hwang, Albert Roach and David J Bjorkman.1AstraZeneca LP, Wayne, PA, United States; 2AstraZeneca LP, Wayne,PA, United States; 3AstraZeneca LP, Wayne, PA, United States; and4University of Utah School of Medicine, Salt Lake City, UT, UnitedStates.

Purpose: Number needed to treat (NNT) provides a meaningful measure ofthe magnitude of difference between 2 therapies and allows a basis for costcomparisons. Using evidence-based medicine concepts, we determined therelative efficacy of esomeprazole versus omeprazole for preventing treat-ment failures (unhealed EE) during acute treatment.Methods: We calculated the NNT to prevent 1 patient with endoscopicallydocumented EE from having unhealed EE after 8 weeks of treatment foresomeprazole 40 mg (n � 2446) relative to omeprazole 20 mg (n � 2431)using pooled healing rate data from 3 large trials that had essentially thesame study protocols. The primary endpoint, endoscopic healing at week 8,was analyzed using a log rank test, and 8-week healing rates were estimatedby life table methods. The relative risk reduction (RRR) and absolute riskreduction (ARR) for treatment failure (no healing) were first determined.The NNT was then calculated as 1/ARR.Results: Pooled response rates and NNTs for esomeprazole versus ome-prazole are shown in the table. In the pooled analysis, esomeprazoleshowed a significantly higher 8-week healing rate than omeprazole (p �0.001). The analysis indicates that 14 patients would need to be treated for8 weeks with esomeprazole 40 mg instead of omeprazole 20 mg to prevent1 treatment failure.Conclusions: Esomeprazole is more effective than omeprazole for healingEE, with an NNT of 14. This NNT can be used as a basis for comparingrelative efficacy and costs of esomeprazole and other PPIs across clinicaltrials.

Eso 40 mg(n � 2446)% Healed(95% CI)

Ome 20 mg(n � 2431)% Healed(95% CI) RRR ARR NNT

93.4%(92.4%–94.5%)

86.2%(84.8%–87.7%)

52.2%(43.5%–60.7%)

7.2%(5.5%–8.9%)

14(11–18)

872

Cost effectiveness of misoprostol vs. omeprazole in the prevention ofNSAID associated gastrointestinal symptoms in high risk patientsLoughney TM, Jai EM, Goo E, Dydek GJ, Beaudoin D, Hammond S,Egan JE, Yoshinobu BH. Tripler Army Medical Center, Honolulu, HI.

Purpose: The objective of this study was to compare the efficacy and totalutilization of resources in a cost-effectiveness analysis of misoprostolversus omeprazole in the prevention of NSAID-associated symptoms inhigh-risk patients.

S273AJG – September, Suppl., 2001 Abstracts