Multiplevariate Logistic Analysis And Model Establishment Of Risk Factors For Inrinotecan-induced Adverse Reactions

Purpose:To explore the main risk factors inducing the severe toxicities of irinotecan and establish Logistic regression models to predict the incidence rate of these toxicities. Evaluate the accuracy of each model of severe toxicity by ROC curve.Methods:we did a retrospective analysis between84patients treated with irinotecan-containing chemotherapy in Qilu Hospital of Shandong University from September2011to September2013. We monitored all adverse reactions after each cycle of chemotherapy, classified according to the3th edition of common terminology criteria for adverse reactions. The serious adverse reactions we paid attention to were severe leukopenia, neutropenia, delayed diarrhea, vomiting and cholinergic responses. For each patient, a detailed record of a total of16factors that might affecting serious adverse reactions of irinotecan was noted:age, gender, radiotherapy condition, operation condition, cance stage, TBIL baseline level before treatment, HGB level before treatment, genotypes of UGT1A1*28and UGT1A1*6, the number of metastasis, cancer location, KPS score, chemotherapy scheme, previous chemotherapy cycles, dose intensity and body surface area. SPSS17.0software was used for statistical analysis of all data. Each kind of adverse reaction respectively did single factor analysis with the16factors for preliminary screening. Factors with p<0.8in the single factor analysis were analyzed by Logistic regression analysis. Finally established the Logistic regression model and evaluated the prediction accuracy by ROC curve.Results:Three factors got into the model equation of severe neutropenia:the genotypes of UGT1A1*28and UGT1A1*6and chemotherapy scheme; correct rate of prediction was81.4%. Five factors got into the model of severe leukopenia in the first cycle:the genotypes of UGT1A1*28and UGT1A1*6, previous chemotherapy cycles, HGB level before treatment and the cancer location (digestive tract or other parts); correct rate of prediction was92.3%. Three factors got into the model equation of severe neutropenia:the genotypes of UGT1A1*28and UGT1A1*6and operation treatment; correct rate of prediction was77.1%. Three factors got into the he model equation of severe neutropenia in the first cycle:genotypes of UGT1A1*28and UGT1A1*6and the cancer location (digestive tract or other parts); correct rate of prediction was75.5%. Only the genotype of UGT1A1*6got into the model equations of severe delayed diarrhea and severe delayed diarrhea in the first cycle; the respective prediction accuracy was69%and73.5%. The TBIL before treatment and gender got into the model equation of cholinergic response; correct rate of prediction was78.1%.Conclusion:For patients of using irinotecan containing chemotherapy, genotypes of UGT1A1*28and UGT1A1*6are critical to irinotecan-induced severe leukopenia and neutropenia; mutant UGT1A1genotype would have higher incidence rate of severe hematologic toxicity. And the homozygous mutation of UGT1A1*6caused more frequent severe neutropenia than the heterozygous mutation. Delayed diarrhea was dependent on the gene type of UGT1A1*6. Women patients with TBIL>15umol/L before treatment would be more prone to occur the cholinergic response.