Font Size: a A A

Multivariate Cox Regression Analysis Of Prognostic Factors Impacting On Survival Of Prostate Cancer And Logistic Regression Modelling For Selecting Biopsy Cores

Posted on:2012-05-26Degree:MasterType:Thesis
Country:ChinaCandidate:S J ZhaoFull Text:PDF
GTID:2214330341952306Subject:Surgery
Abstract/Summary:PDF Full Text Request
ObjectiveWe investigate the prognostic factors for prostate cancer(Pca)and find out variables which closely related to the survival of Pca.MethodsThe information of 431 patients of Pca in Guangzhou First Municipal People's Hospital from 2001 to 2010 were retrospectively analysed. They were followed up, and then their survival status and prognostic factors were analysed. The correlation between clinicopathological parameters and immunohistochemical markers were analysed by Chi-square test. We used Kaplan-Meier method for survival function analysis and drew their cumulative survival curves and Log-rank method for significance test. Use univariate and multivariate Cox regression to screen independent prognostic factors for Pca.Results(1) The last follow-up date was 31th December 2010. At the final follow-up, 190 cases were alive, 108 cases were died and 43 cases were lost to followed up. The median overall survival time was 82.00 (58.67~105.33) months, and overall survival rates at l-, 3- and 5-year were 90.3%, 57.4%, 36.8%, corresponding disease specific survival rates were 91.1%, 60.7%, 40.0%. (2)Univariate COX regression analysis of clinicopathological group: Percentage of Positive Cores(PPCORE)(HR=4.16,P=0.003),Gleason Score(GS)(HR=1.24,P=0.02) and distant metastasis(HR=1.67,P<0.001)were risk factors of survival of prostate cancer,in which HR of PPCORE was highest.(3)Multivariate COX regression analysis of clinicopathological group: metastasis (HR=1.67, P=0.002) was the main risk factor impacting on the survival of Pca. There was significantly difference of survival rates among no metastasis group, M1a group and≥M1b group (X2 = 39.60, P<0.001). Their average survival times were 79.52±3.92 (CI71.83~87.21) months, 53.61±9.53 (CI34.93~72.28) months and 44.69±4.36 (CI36.13~53.24) months.(4)Immunohistochemical markers group: Univariate COX regression analysis resulted that Ki-67 expression(HR=1.49, P=0.002) indicates poorer prognosis. Chi-square test showed that Ki-67 expression related to GS﹥7 (P<0.001). PSA, PSAP, P63,α34βE12 and GS did not related to GS﹥7(P﹥0.05).Conclusion(1)Higher GS, higher PPCORE, and distant metastasis indicate poorer prognosis of Pca, in which PPCORE has biggest influence on the survival.(2)Metastasis was the main risk factor impacting on the survival of Pca.(3)We could suggest Pca patients of Ki-67 positive expression to take active treatment.Part 2. Comparison in schemes of prostate biopsy of different biopsy coresObjectiveThe study is to compare positive rates and PPCOREs in different schemes of prostate biopsy and investigate how to select the biopsy scheme.MethodsRetrospective analysis of the clinical data in Guangzhou First Municipal People's Hospital of total prostate specific antigen (T-PSA)≥4ng/ml and received TRUS-guided prostate biopsy during 2001 to 2010, 409 cases of which received 13-core systemic biopsy, 19 cases of which received 6-core systemic biopsy, and assuming that reducing the 3 central cores of 13-point systematic prostate biopsy is 10-core systematic biopsy. We compared postoperational complication rates and PPCORE in different schemes, the positive rates in different groups classified by PSA levels and different prostate volumes.ResultsPostoperational hematuria rates of 13-core biopsy group and 6-core biopsy group were 53.3% and 15.8%, which were significantly different(P<0.01);The incidence rates of bloody faeces, infection, pain, vasovagal reflex between the above two groups were not statistically different (P﹥0.05). The positive rates between 13-core biopsy group and 10-core biopsy group were not statistically different (P﹥0.05). In the same biopsy scheme, positive rates of different levels of T-PSA biopsy were significantly different (P<0.01). In the same levels of T-PSA, positive rates between 13-core biopsy and 10-core biopsy were not statistically different (P﹥0.05). In the same biopsy scheme, positive rates of different prostate volume were significantly different (P<0.01). In the same prostate volume, positive rates between 13-core biopsy and 10-core biopsy were not statistically different (P﹥0.05).The mean PPCORE of 10-core biopsy and 13-core biopsy were 0.282±0.380 and 0.286±0.382, they were not significantly different(P=0.865).Conclusion:(1) The positive rates and PPCOREs between 10-core biopsy and 13-core biopsy were not significantly different, and postoperational hematuria rates of 10-core biopsy significantly reduced. (2)The prostate cancer detection rate increased gradually with the increase of PSA. The prostate cancer detection rate decreased gradually with the increase of prostate volumn. (3)When PSA﹥100ng/ml, if prostate volume<40ml we could select 6-core biopsy,and if prostate volume﹥40ml we could select 10-core biopsy. When PSA<100ng/ml, if prostate volume<40 ml we could select 10-core biopsy, and if prostate volume﹥40ml we may consider increasing the biopsy core number, such as 18, 21, 24-cores and so on. ObjectiveThe objective is to study correlation of PPCORE in prostate biopsy and develop logistic models for predicting biopsy core number.Methods451 subjects in Guangzhou First Municipal People's Hospital from 2001 to 2010 underwent TRUS-guided transrectal biopsy with 13-core method screening for prostate carcinoma. Their clinical datas were retrospectively analysed. The following variables were recorded in each patient: age, total prostate specific antigen (PSA), free to total PSA ratio (F/TPSA), prostate volume(PV), transition zone volume(TZ), PSA density(PSAD), PSA-TZ density (PSATZ), (F/T PSA)/PSAD, digital rectal examination(DRE), and TRUS findings. Compare positive rates of each core of 13-core biopsy and each part of prostate by one-way ANOVA analysis. Use Spearman coefficients of correlation between PPCORE and each variable. Make linear regression equation by PPCORE as dependent and all variable as independents. Compare the difference of each variable in the two PPCORE groups by two-independent-samples nonparametric test. Use multivariate logistic regression analysis and ROC curves to build models that predict biopsy positive result and PPCORE≥0.5,then we can get the predicted value P. P<0.05 is considered as significant difference.Results(1)The median age and serum PSA were 72 years and 19 ng/ml. Prostate biopsy was positive for cancer in 210(46.6%) patients. Mean PPCORE in cohort is 0.30±0.38. The positive rates of each core were not statistically different (P﹥0.05).(2)Multivariate logistic regression analysis that predict biopsy positive result: Higher PSAD(OR= 12.606, P<0.001)and positive DRE result (OR =3.067, P<0.001) indicate greater possibility of biopsy positive result (OR =22.25, P=0.001). The logistic regression model that predict biopsy positive result was logit P=-8.311+0.083age+1.121DRE+2.534PSAD, its sensitivity and specificity were 0.803 and 0.773. The cutoff value P was 0.343. When P≥0.343, it was considered as possibly Pca.(3)TPSA, PSAD, PSATZ and (F/T PSA)/PSAD significantly correlated with PPCORE (r=0.52, 0.57, 0.57 and -0.55). All variables correlated with biopsy positive result(P<0.001). TPSA, F/T PSA, PSAD, PSATZ and(F/T PSA)/PSAD closely correlated with PPCORE≥0.5(P=0.024~<0.001).(4)DRE was a good variable that predicted PPCORE.(5)The logistic regression model that predicts PPCORE≥0.5 was logit P = 0.042-0.003*TPSA+0.176*PSATZ, its sensitivity and specificity were 0.612 and 0.703. The cutoff value P was 0.595. When P≥0.595, it was considered as possibly PPCORE≥0.5. When P<0.595, it was considered as possibly PPCORE<0.5.Conclusion(1)It is hard to select special biopsy cores for important biopsy cores.(2)We should notice DRE clinically.(3)In the logistic regression model logit P=-8.311+0.083age+ 1.121DRE+ 2.534PSAD, we could get predicted value P. When P≥0.343, it was considered as possibly Pca, then we could suggest patients to undergo prostate biopsy.(4)Based on the 10-core biopsy scheme, we use logistic regression model logit P = 0.042-0.003*TPSA+0.176*PSATZ, then we could get predicted value P. When P≥0.595, it was considered as possibly PPCORE≥0.5,then we could consider decreasing biopsy cores. When P<0.595, it was considered as possibly PPCORE<0.5, then we could consider increasing biopsy cores, for example,18, 21, 24 cores and so on. The specific biopsy cores number accordances with the clinician's experience on prostate biopsy.
Keywords/Search Tags:Prostate Cancer, Biopsy cores, Cox regression analysis, Logistic regression model
PDF Full Text Request
Related items