Font Size: a A A

The Application Of Time-dependent ROC Curves In The Prognosis Of Non-Hodgkin’s Lymphoma

Posted on:2013-01-14Degree:MasterType:Thesis
Country:ChinaCandidate:L WuFull Text:PDF
GTID:2284330371978978Subject:Epidemiology and Health Statistics
Abstract/Summary:PDF Full Text Request
Objective To comprehensively and quantitatively assess the prognostic value of beta2-microglobulin(β2-MG) in non-Hodgkin’s Lymphoma(NHL) via meta-analysis. Then, constructrd two Cox regression models, one only included IPI,the other included IPI and β2-MG. Time-dependent reciever operating characteristic (ROC) curves method was used to evaluate the two Cox regression models’prediction accuracy. We suggested adding β2-MG into IPI in order to offer the theory support for the clinical practice.Methods Pubmed, Science Direct and OVID in English from1980to2012were all searched to extract relevant information using predefined retrieval strategy by two assessors. The hazard ratio (HR) estimates of overall survival (OS) between elevated and normal groups of β2-MG were derived or calculated. Fixed and random effects models were successively used to synthesize these HRs. funnel plot and fail safe number(Nfs) were used to estimate the publication bias. Two standard Cox regression model outputs were used to obtain estimates of time-dependent sensitivity and specificity, and time-dependent reciever operating characteristic (ROC) curves. Time-specific area under the ROC curves(AUC(t)) measured the probability that the marker value for a randomly selected decedent exceeds the marker value for a randomly selected survivor. Two global concordance measure (C index) were used to compare the whole follow up predictive accuracy of the two models.Results A total of17articles were included in meta-analysis. The cumulative patients with NHL were2402. Homogeneity test showed that heterogeneity between studies exists (I2=82%, P<0.001).The pooled hazard ratio was2.70(95%CI1.90-3.83) and the overall effect were confirmed to be statistically significant (Z=5.57, P<0.001). Subgroup and sensitivity analysis also revealed β2-MG was an important hazard factor to NHL patients. Four hundred twenty-nine NHL patients were followed up from2008to2011in Shanxi tumor hospital, the median survival time was36months. Time-specific AUC of two models decreased gently in45months. The global concordance measures were0.671(SE=0.0149,95%CI0.638-0.687) and0.704(SE=0.0150,95%CI0.674-0.723). There was statistical significance between two groups of model score(Z=3.254,P<0.001).Conclusion Combined hazard ratio suggested that the mortality risk of elevated β2-MG was2.7times as higher as the normal one. β2-MG may be an useful independent prognostic factor for NHL. It’s helpful for clinical decision to add β2-MG into IPI.
Keywords/Search Tags:β2-MG, IPI, non-Hodgkin’s Lymphoma, meta-analysis, ROC curves
PDF Full Text Request
Related items