Frontiers in Cancer Prevention Research

ors. Furthermore, a small proportion (3%) of data generatedfrom medical records based on community gastroenterologists andpathologists are prone to difficulty in generating accurate endpointdata and will require further review prior to analysis. #C84 Biomarkers for Validation of Dietary Exposure Assessments inCancer Epidemiology. Natasa Tasevska, Shirley A. Runswick, Sheila A.Bingham. MRC Dunn Human Nutrition Unit, Cambridge, England.Environmental factors, especially diet, are thought to be major con-tributors to human cancer. However, it has recently been shown, usingbiomarkers, that epidemiological methods to measure diet are associ-ated with large correlated measurement error therefore affecting thereliability and interpretation of the studies investigating the associa-tion between diet and cancer (Int J Epid 2001, 30, 309-317; PublicHealth Nutr 2002, 5 (6A), 915-923). Only limited biomarkers to assessaccuracy of dietary intake are available and there are many foods andnutrients for which no suitable biomarker has been identified (PublicHealth Nutr 2002, 5, 821-827). We have investigated two new biomark-ers of nutrients widespread in food, thiamine and sugars. After a doseresponse of sugars and thiamine in 24-h urine to increased levels infood was found, the relation between these markers in food and urinewhile subjects consumed their normal varying diet was investigated.Thirteen volunteers (6 women, 7 men), living in the MRC Dunn HumanNutrition Unit Volunteer Suite, were given their normal varying diet(replicated from food diaries kept beforehand) for 30 days andamounts consumed assessed by weighing dishes before and after con-sumption. The subjects collected 24-h urine on a daily basis through-out the study, verified for completion with PABA (Clin Sci 1983, 64,629-635). Sucrose and fructose, and thiamine in urine were measuredas markers for sugars and thiamine intake, respectively (Internat J VitNutr Res 1986, 56, 155-159; in Methods of Enzymatic analysis, 2nd edition 1974). Intakes were calculated using the DINER software (PublicHealth Nutr 2002, 4, 1253-1265). Average intake of total sugars was 201.9±68.8 g/d (within-person CV=21.5%) and average excretion inurine was 36.5±16.6 mg/d (CVwp=50.5%) for sucrose and 61.8±61.3 mg/d (CVwp=38.6%) for fructose. The correlation between total sugars intake and the sum of sucrose and fructose excretion levels in urinewas r=0.841 (p<0.001). In 8 subjects, average intake of total thiaminewas 2.2±0.5 mg/d (CVwp=27.0%) and average excretion in urine was 494.8±166.1μg/d (CVwp=27.0%). The correlation between thiamine indiet and levels of thiamine in urine was r=0.69 (p=0.058) and results ofthe complete analysis will be presented. High within-subject variability identified in the daily excretion level of both potential markersimplies that substantial number of urine collections might be neededto gain an acceptable level of precision in validation of sugars and thiamine intake on an individual level, but a single urine collection maybe valuable comparative/calibration marker of their intake on a popu-lation basis. (This research is supported by the MRC, the British Counciland WCRF.). #C85 Use of Cluster Analysis to Define Low Levels of MercaptopurineMetabolites in Adolescents with Acute Lymphoblastic Leukemia: HowLow is “Low”? Mary Ann O’Riordan, Fatoumata Traore, CarolynMyers, Karen Groth, Ahna Hoff, Dennis Drotar, Eric Kodish. CaseWestern Reserve University SOM, Cleveland, OH. Introduction: Oral mercaptopurine (6-MP) is an important component of standard maintenance therapy for childhood acute lym-phoblastic leukemia (ALL) (Relling et al., 1999). Its therapeutic effect depends upon conversion to cytotoxic thioguanine nucleotides (TGN)that can disrupt DNA structure (Lennard and Lilleyman, 1989). A second metabolic pathway, catalyzed by the enzyme thiopurine methyltransferase (TPMT), leads to production of methylated mercaptop-urines (MMP). Low concentrations of TGN have been shown to be asso-ciated with poor disease outcome in children with ALL (Lennard andLilleyman, 1989; 1994). Few studies have attempted to correlate treat-ment outcome with RBC MMP levels, but there has been a suggestionthat this metabolite may also have activity against leukemia cells (Erbet al., 1998). While many underlying causes may result in less than optimal TGN and MMP RBC concentration levels, the potential for a poorclinical outcome is the same. Thus, the ability to screen patients forrelapse risk based on their metabolite level is of considerable clinicalimportance. Objective: To use standard statistical techniques to grouppatients receiving treatment for acute lymphoblastic leukemia (ALL)according to their oral mercaptopurine (6-MP) metabolite levels and toestablish cut-off points to screen for potential poor clinical outcome.Study design. This is a methodological study using 81 observationsfrom 27 adolescent ALL patients enrolled in a multi-site adherencestudy. Cluster analysis was the primary analytical technique. Results.The mean age for the group was 16 years (SD 2), and 74.1% weremales; 41% were members of a minority racial/ethnic subgroup. Fourclusters were retained as our optimal solution (Figure 1). Cluster 1 wascharacterized by a high TGN level and a low MMP level and cluster 2 ischaracterized by a low TGN and a high MMP level. Cluster 4 was char-acterized by a low TGN and a very high MMP level. These clustersexhibit the expected negative correlations between the two metabo-lites. In contrast, Cluster 3 had both low TGN and low MMP levels. Fiveout of the 27 adolescents had documented discontinuous 6-MP thera-py. They grouped in Cluster 3 during the time that their medicationswere stopped, but grouped in other clusters at other times. The medi-an ANC was highest in Cluster 3, consistent with low drug concentra-tions (Kruskal-Wallis overall p-value= 0.12). Values of both TGN andMMP at or below the respective 95th percentile for values in Cluster 3were selected as the definition for higher risk of relapse.Conclusion.This study provides an objective method for identifying patients poten-tially at risk for treatment failure due to suboptimal drug therapy. #C86 Cigarette Smoking, NAT1 and NAT2, and the Risk of AdvancedColorectal Adenoma. Roxana Moslehi,1 Jinbo Chen,1 David W. Hein,2Nilanjan Chatterjee,1 Timothy R. Church,3 Joel Weissfeld,4 Richard B.Hayes.1 National Cancer Institute,1 Rockville, MD, University ofLouisville,2 Louisville, KY, University of Minnesota,3 Minneapolis, MN,University of Pittsburgh,4 Pittsburgh, PA.Cigarette use is associated with risk for colorectal adenoma, a col-orectal cancer precursor. N-acetyltransferases, NAT1 and NAT2, are important enzymes for detoxification of certain aromatic amine car-cinogens and, following N-hydroxylation, the activation of otheramine protocarcinogens to their ultimate carcinogenic form.Polymorphisms in the coding regions of NAT1 and NAT2 may alterenzymatic activity and modify cancer risks related to exposures tothese carcinogens. In vitro studies have characterized the acetylatorphenotypes associated with various NAT2 diplotypes as slow, intermediate and rapid. In the Prostate, Lung, Colorectal and Ovarian (PLCO)Cancer Screening Trial, we compared NAT1 and NAT2 gene variant distributions for 772 cases with left-sided advanced adenoma (size >1cm,villous structure, or high-grade dysplasia) and 777 gender and age-matched controls screened negative for cancer, adenomas and otherprecursor lesions. All analyses were adjusted for gender, race and ageat interview. Risks for advanced adenoma were significantly increased among recent smokers (current smokers or those who quit< 10 yearsago) (OR=2.3, 95%CI: 1.7-3.1) and among those who smoked morethan 20 cigarettes per day (OR=1.7, 95%CI: 1.3-2.2), compared to non-smokers. Among the controls, genotype distributions of the polymor-phic loci studied within NAT1 (T1088A, C1095A, C-344T, A-40T) andNAT2 (A803G, G857A, C282T, T341C, C481T, and G590A) were in Hardy-Weinberg equilibrium. Haploview analysis revealed strong linkage disequilibrium within the two genes, but not between the genes. NAT1and NAT2 haplotypes and diplotypes were created and acetylator phe-notypes for NAT2 were deduced from the diplotype data. The mostcommon NAT2 diplotype among the controls was NAT2*4/*5B (17.8%)and among the cases was NAT2*5B/*6A (20.1%). Trend analysis