Comparison Between Serum HE4,Ca125 And On Which Roma,CPH-I Based In Differentiating Malignant From Benign Ovarian Tumors

OBJECTIVETo compare diagnostic value of HE4,CA125 and on which risk of ovarian malignancy algorithm(ROMA),Copenhagen index(CPH-I)based in differentiating malignant from benign ovarian tumors.METHODSData of 719 consecutive patients of ovarian tumors,who visited Department of Obstetrics and Gynecology,Nanfang Hospital of Southern Medical University from Sep 2014 to Nov 2016 were analyzed retrospectively,and according to the research demands,patients were divided into different groups,including 119 cases with epithelial ovarian cancer(EOC),25 cases with non-epithelial ovarian cancer(Non-EOC),44 cases with borderline ovarian tumors(BOT),531 cases with benign ovarian tumors or 96 cases with non-less common ovarian histopathologies(Non-LCOHs),92 cases with less common ovarian histopathologies(LCOHs),531 cases with benign ovarian tumors.Among Non-LCOHs,4 cases with metastatic ovarian cancer were included for its poor prognosis.In our study,the skewed distribution measurement data were expressed as Median(P25,P75),enumeration data as numerical value and percentage.For non-normal distribution measurement data and ranked data,Kruskal-Wallis Test were used for multiple comparisons and Mann-Whitney Test for pairwise comparison.For frequency of categorical data,Chi-Square Test were used.Areas Under Curve(AUC)of receiver operating characteristic(ROC)curve were calculated based on golden standard of pathologic diagnosis,and the others index like sensitivity,specificity,positive predictive value,and negative predictive value were obtained via referring to corresponding cut-off values and chi-square test,in order to compare diagnostic value of HE4,CA125 and on which ROMA,CPH-I based in differentiating malignant from benign ovarian tumors.All statistical analyses were performed with SPSS(version 20.0)software,and P<0.05 was regarded as statistically significant.RESULTS1.The age and rate of postmenstrual cases in patients with EOC,Non-EOC,BOT or Non-LCOHs,LCOHs were greater than those with benign tumors,and differences between each group were significant except Non-EOC and benign group,Non-EOC and BOT group.Among ovarian cancer,68.9%patients with EOC were staged advanced,while those with Non-EOC or BOT,68.0%and 95.5%were early stage,respectively,proportion of advanced patients in EOC group was higher than another two groups significantly.When malignant patients were divided into Non-LCOHs and LCOHs,majority of the former one was advanced patients(77.1%),while the latter one early stage patients(80.4%),and they differred significantly from each other.2.When patients were segmented into EOC,Non-EOC,BOT,benign groups,comparison between any two groups showed that level of serum HE4,CA125,ROMA,CPH-I between each other were different significantly other than Non-EOC and BOT groups,and the order were as follows:EOC group>Non-EOC group>BOT group>benign group.Similarly,significant differences were observed when pairwise comparison were performed between Non-LCOHs,LCOHs and benign groups and level in Non-LCOHs group were higher,benign group lower.3.CPH-I had the most maximum AUC than others,ROMA and HE4 tied for second and CA125 the most minimum in differentiating malignant from benign ovarian cancer patients(AUC of CPH-I was the highest,followed by ROMA,HE4,CA125 in premenstrual or advanced stage population and HE4 followed by CPH-I,ROMA,CA125 in postmenstrual population and ROMA followed by HE4,CPH-I,CA125 in early stage population).CA125 made highest sensitivity and ROMA secondary,CPH-I third,HE4 last in whole population(CA125 change into third in postmenstrual or advanced stage population and ROMA was equal to CPH-I in former population and higher than CPH-I in latter one).However,the order of specificity of different indexs was contrary to that of sensitivity.4.When discriminating between malignant and benign patients by HE4,CA125,ROMA,CPH-I accordingly,the number and ratio of false negative cases for EOC group were 28(23.5%),19(16.0%),15(12.6%),18(15.1%),and 20(80.0%),12(48.0%),15(60.0%),17(68.0%)for Non-EOC groups,and 38(86.4%),25(56.8%),28(63.6%),33(75.0%)for BOT group or 15(15.6%),8(8.3%),7(7.3%),6(6.2%)for Non-LCOHs group,and 71(77.2%),48(52.2%),51(55.4%),62(67.4%)for LCOHs group.The number and ratio of false positive cases for benign group were 6(1.1%),144(27.1%),66(12.4%),28(5.3%).5.When differentiating EOC from benign one,AUC of ROMA was highest,followed by HE4,CPH-I,CA125 successively(ROMA was equal to CPH-I and followed by HE4,CA125 in premenstrual or advanced stage population,and HE4 followed by ROMA,CPH-I,CA125 in postmenstrual or early stage population).As for sensitivity,the results from highest to lowest,were list as order:ROMA,CPH-I,CA125,HE4(CPH-I was exceeded by CA125 in premenstrual or early stage population and was equal to ROMA in postmenstrual population).However,HE4 had the highest specificity followed by CPH-I,ROMA,CA125 as previously described.When differentiating Non-LCOHs from benign cases,ROMA had the most maximum AUC,CPH-I and HE4 tied for second,CA125 the most minimum(CPH-I followed by ROMA,HE4,CA125 in premenstrual or advanced stage population,and HE4 followed by ROMA,CPH-I and CA125).CPH-I had the highest sensitivity,and ROMA,CA125,HE4 one after another(CA125 was equal to the highest in premenstrual population and to ROMA in advanced stage population,and ROMA was equal to the highest in postmenstrual or early stage population).Specificity was the same as previously described.CONCLUSION1.Overall,HE4,ROMA,and CPH-I did the comparative performance in differentiating malignant from benign ovarian tumors and both were better than CA125;2.In terms of predicting malignant patients,ROMA and CPH-I were inferior to CA125 when all OC patients were included(it was mainly observed in premenstrual or early stage population,while both three models were done similarly in.postmenstrual or advanced stage population),and superior to that when only Non-EOC or Non-LCOHs were included.Moreover,when only Non-LCOHs population,of whom were with poor prognosis,were included,CPH-I was superior to ROMA slightly.Meanwhile,as element of CPH-I,age was an easy obtainable and reproducible variable when comparing to menstrual status included in ROMA,that is to say CPH-I was more simple and useful than ROMA.Both of these models performed better than HE4;3.As a novel tumor biomarker,HE4 had the best specificity for predicting benign ovarian diseases,followed by CPH-I,ROMA,CA125 respectively and the three former made up for shortcoming of weak specificity of CA125;4.However,in common with Non-EOC,BOT patients or LCOHs patients,the rate of missed diagnosis in premenstrual or early stage population were still staying at a high level and further research and exploration for more effective differential diagnosis tools were needed.