Standard and optimal cut-off values of serum ca-125, HE4 and ROMA in preoperative prediction of ovarian cancer in Vietnam.

Among the 277 patients included in the study, 247 (89.2%) were diagnosed with benign ovarian tumors and 30 (10.8%) were diagnosed with EOC. The differences in age and menopausal status were statistically significant, in the cancer group, 63.3% of the patients were older than 50 years, and 56.7% of patients were already menopausal ( p  < .01). The numbers of children were not significantly different between the two groups ( p  = .220) ( Table 1 ). Table 1 Demographic characteristics of study's subjects. Table 1 Benign tumor EOC p n (247) % n (30) % Age <20 16 6.5 0 0 <0.01 20–29 76 30.8 3 10.0 30–39 66 26.7 2 6.7 40–49 52 21.1 6 20.0 ≥ 50 37 15.0 19 63.3 

 Menopausal status Pre-menopausal 217 87.9 13 43.3 <0.01 Post-menopausal 30 12.1 17 56.7 

 Number of children Nullipara 70 28.3 7 23.3 0.220 1 48 19.4 4 13.3 2 52 21.1 4 13.3 ≥ 3 77 31.2 15 50.0 Demographic characteristics of study's subjects. Median values of HE4, CA125 and ROMA of the cancer group were statistically higher than those of the benign tumor group (Mann-Whitney test) ( Table 2 ). The median value of HE4 in the EOC group was 90.00 pmol/l (57.47–447.97 pmol/l), which was statistically higher than the value from the benign tumor group at 38.50 pmol/l (30.10–45.90 pmol/l) ( p  < .01). The median value of CA-125 of the EOC group was 214.20 U/ml, (60.56–764.42 U/ml), significantly higher than the value of the benign tumor group at 17.45 U/ml (11.88–28.70 U/ml) (p < .01). The ROMA median value in the benign tumor group was 4.47% (2.44% - 7.02%), and that of the EOC group was 55.20% (11.89–95.0%); the difference was statistically significant (p < .01). Table 2 Values of serum CA125, HE4, and ROMA of study's subjects. Table 2 Median (Q25%-Q75%) Study sample ( n  = 277) Benign tumor ( n  = 247) EOC ( n  = 30) Z value p CA125(U/ml) 18.91 (12.20–39.04) 17.45 (11.88–28.70) 214.20 (60.56–764.42) −6.647 <0.01 HE4(pmol/l) 39.80 (30.50–50.10) 38.50 (30.10–45.90) 90.00 (57.47–447.97) −7.040 < 0.01 ROMA (%) 4.80 (2.53–8.47) 4.47 (2.44–7.02) 55.20 (11.89–95.00) −7.373 < 0.01 Values of serum CA125, HE4, and ROMA of study's subjects. ROMA yielded a higher AUC value than those from CA-125 or HE4 (see Table 3 ), both in the general group as well as the post-menopausal group, but the differences were not statistically significant (Hanley-McNeil test): ROMA vs. CA125 (Z = 0.869; p  = .3851); and ROMA vs. HE4 (Z = 1.090, p  = .27) (see Graph 1 , Graph 2 ). With an optimal cut-off value of 9.52% for ROMA, the sensitivity and specificity were 86.7% and 88.7%, respectively. In both the pre- and postmenopausal groups, optimal cut-off values of ROMA yielded high specificity and negative predictive values. Table 3 Validity of CA 125, HE4, ROMA for preoperative diagnosis of ovarian cancer at standard and optimal cut-offs. Table 3 ROC Optimal cut-off Standard cut-off Cut-off Se (%) Sp (%) PPV NPV Cut-off Se (%) Sp (%) PPV NPV Study group ( n  = 277) CA125 0.872 49.4 U/ml 83.3 86.6 43.1 97.7 35 U/ml 83.3 78.5 31.65 97.47 HE4 0.894 55.4 pmol/l 80.0 91.5 53.3 97.4 – 50.0 98.38 78.95 94.19 ROMA 0.912 9.52% 86.7 88.7 48.1 98.2 – 80.0 84.6 38.71 97.21 

 Pre-menopausal ( n  = 230) CA125 0.931 49.4 U/ml 92.3 85.7 27.9 99.5 35 U/ml 92.3 77.0 19.05 99.40 HE4 0.835 55.4 pmol/l 76.9 94.0 43.5 98.6 70 pmol/l 61.5 98.2 66.67 97.71 ROMA 0.844 9.89% 76.9 92.6 38.5 98.5 7.4% 76.9 82.9 21.28 98.36 

 Post-menopausal ( n  = 47) CA125 0.900 48.9 U/ml 76.5 93.3 86.7 87.5 35 U/ml 76.5 90.0 81.25 87.10 HE4 0.894 59.3 pmol/l 76.5 86.7 76.5 86.7 140 pmol/l 41.2 96.7 100.0 75.0 ROMA 0.924 25.62% 82.4 96.7 93.3 90.6 25.3% 82.4 96.7 93.33 90.62 Graph 1 ROCs of CA125, HE4, and ROMA values in pre-menopausal women. Graph 1 Graph 2 ROCs of CA125, HE4, and ROMA values in post-menopausal women. Graph 2 Validity of CA 125, HE4, ROMA for preoperative diagnosis of ovarian cancer at standard and optimal cut-offs. ROCs of CA125, HE4, and ROMA values in pre-menopausal women. ROCs of CA125, HE4, and ROMA values in post-menopausal women.

This study was conducted to evaluate the validity of serum CA-125, HE4 and ROMA at standard and optimal cut-offs in the preoperative prediction of EOC in Vietnam. Using a cut-off value of 35 U/ml, the sensitivity of CA-125 in our study was 83.3% and the specificity was 78.5%. The area under the ROC curve of CA-125 was 0.872; with the optimal cut-off of 49.4 U/ml, the sensitivity was 83.3% and the specificity was 86.6%. The level of serum CA-125 can be increased in many clinical cases with benign conditions, including pregnancy, endometriosis, uterine fibroids, pancreatitis, menstruation, pelvic inflammatory disease, and liver disease. On the other hand, serum CA-125 is not elevated in approximately 20% of women with ovarian cancer. In our study, cases of endometriosis, ovarian cysts and inflammation showed very high CA-125 levels. The sensitivity of CA-125 in distinguishing benign from malignant tumors ranged between 61% and 90%, while specificity ranged between 35% and 91% ( American College of Obstetricians and Gynecologists, 2007 ). The positive predictive value (PPV) in women with an adnexal mass ranged from 35% to 91%, and the negative predictive value ranged from 67% and 90% ( American College of Obstetricians and Gynecologists, 2007 ). The median value of HE4 in the EOC group was 90.00 (57.47–447.97) pmol/l, which was statistically higher than the value in the benign group, at 38.50 (30.10–45.90) pmol/l ( p  < .01). This result confirmed the findings from many previous studies abroad ( Dikmen et al., 2015 ; Fujiwara et al., 2015 ; Van Gorp et al., 2011 ; Karlsen et al., 2015 ). Our HE4 values were similar to those reported by To Thi Thuc Trang in 2014, with the HE4 median of 44.7 pmol/l (36.7–55.2 pmol/l) ( Trang, 2014 ), and differed from those of Sandri, with median of 333.6 pmol/l (25.7–3.013.9 pmol/l) ( Sandri et al., 2013 ) and Jacob et al. (2011) with median for benign tumor group of 50 pmol/l (42–62 pmol/l), and for EOC group of 128 pmol/l (79–572 pmol/l) ( Jacob et al., 2011 ). This difference could be due to the different sample sizes and possibly the nature of non-Asian population. Our study was carried out on a population of women with ovarian tumors hospitalized for surgery, using pathologic results as the gold standard for the prediction of EOC, based on cut-off values of HE4 > 140 pmol/l in postmenopausal women, and > 70 pmol/l in pre-menopausal women. The sensitivity of HE4 in our study was 50% and the specificity was 98.38%. Compared to the range of sensitivity of 56.9% - 80.8% from previous studies abroad, our results were lower. The specificity of HE4 in our study was 98.38% and was consistent with national and international studies, most of which were ≥ 90%: the study from Trang (2014) yielded specificity of 97.3% ( Trang, 2014 ), while Van Gorp et al. (2011) and Chan et al. (2013) both also achieved a specificity of 96.9%, almost similar to our result ( Van Gorp et al., 2011 ; Chan et al., 2013 ). Only the Jacob study (2011) on 160 patients had a lower specificity at 83.3% ( Jacob et al., 2011 ) (see Table 4 ). Table 4 Diagnostic validity of HE4 from literature. Table 4 Author n Se (%) Sp (%) ROC Van Gorp T. (2011) ( Van Gorp et al., 2011 ) 389 74.5 96.9 0.86 Jacob F. (2011) ( Jacob et al., 2011 ) 160 78.9 83.3 Karlsen M.A. (2012) ( Karlsen et al., 2015 ) 1218 63.2 90 Chan K. K. (2013) ( Chan et al., 2013 ) 414 56.9 96.9 Ba Quyet Vu (2014) ( Ba Quyet Vu, 2014 ) 85 63.6 94.7 Trang TTT (2014) ( Trang, 2014 ) 1290 80.8 97.3 0.92 Wei S.U.(2016) ( Wei et al., 2016 ) 158 75 97.87 0.990 This study Standard cut-off Optimal cut-off 277 50 80.0 98.38 91.5 0.894 Diagnostic validity of HE4 from literature. In Hamed et al. (2013) , when compared with CA-125 as a tool for detecting ovarian cancer, HE4 was more sensitive (90% vs. 83%) and specific (95.0 vs. 85.0%). HE4 also had higher positive (93.1% vs. 80.7%) and negative (92.7% vs. 87.2%) predictive values ( Hamed et al., 2013 ). From our study, the area under the ROC curve (AUC) of HE4 is 0.894, similar to To Thi Thuc Trang's (2014) finding of 0.92 ( Trang, 2014 ). At the optimal cut-off value of HE4 (55.4 pmol/l), sensitivity and specificity for EOC were 80% and 91.5%, respectively. Following the first study on the combination of CA125 and HE4 for improving the diagnostic validity in differentiation between malignant and benign ovarian tumors in 2008 ( Moore et al., 2008 ), Moore et al. (2009) introduced the algorithm known as ROMA (Risk of Ovarian Malignancy Algorithm) to predict ovarian malignancy ( Moore et al., 2009 ). The ROMA median value in the non-ovarian cancer group was 4.47% (2.44% - 7.02%), and the value from the EOC group was 55.20% (11.89–95.0%); this difference was statistically significant ( p  < .01). According to Terlikowska et al. (2016), ROMA's median values of non-cancer and EOC groups were 12.4% and 50.4%, respectively ( Terlikowska et al., 2016 ). ROMA's values reported by Cymbaluk-Ploska et al. (2016) were 11.3% and 96.12% from the benign and EOC groups, respectively ( Cymbaluk-Ploska et al., 2016 ). Explanation for this variation could be the differences in number and rate of EOC cases among these studies. ROMA is the result of an algorithm that combines the values of HE4, CA125 with the consideration of menopausal status, thus calculating the malignancy potential of tumors. The sensitivity and specificity of ROMA at standard cut-offs in our premenopausal were 76.9% and 82.9%; and post-menopausal groups were 82.4%, and 96.7%, respectively. According to Vo Thanh Nhan's study (2010) on 31 cases, the sensitivity of ROMA was 88.2% and the specificity was 64.3% ( Thanh Nhan, 2010 ). Anton et al. (2012) study on 128 patients showed the sensitivity of 74.1% and the specificity of 75.8% ( Anton et al., 2012 ). Su Wei et al.'s study on 158 cases yielded the sensitivity of 93.75%, and the specificity of 92.55% ( Wei et al., 2016 ). According to Ba Quyet Vu (2014) , the sensitivity of ROMA was 63.6% and the specificity was 86.7% ( Ba Quyet Vu, 2014 ). HE4 could be more specific than CA125 in the diagnosis of ovarian cancers. Karlsen et al. recommended the use of ROMA especially among postmenopausal women; this resulted in a more accurate prediction ( Karlsen et al., 2015 ). In the pre-menopausal group, the modified ROMA cut-off value of 9.89% yielded a better specificity in comparison to those of standard cut-off (92.6% vs. 82.9%), while keeping the sensitivity at 76.9%. In the post-menopausal group, since there was no significant difference between optimal and standard cut-off values, diagnostic validity was still unchanged. In, Winarto et al., 2014 reported on the modification of cut-off values of CA-125, HE4, RMI score and ROMA, resulting in higher accuracy compared to the standard ones, at the cost of reduced sensitivity ( Winarto et al., 2014 ). To the best of our knowledge, this is the first study on HE4 and ROMA performed preoperatively to assess and predict epithelial ovarian cancer on Vietnamese patients available to international medical literature. This is of great value, especially if the patient is hospitalized at secondary-level settings and/or in emergency status, where skilled gynecologic surgeons are not universally available. Even at tertiary-level settings, this prediction before actual surgical intervention could improve the management plan for the patient during surgery and afterwards. The strengths of this study include the validated biomarker testing procedures and the standardized pathologic interpretation of tumor specimens. Limitations of present study include the small number of subjects having EOC diagnoses involved in this study, and the nature of targeted hospital based patient population (scheduled for surgery) and not the outpatient based population.

Serum CA 125, HE4 levels and ROMA index have good validity in the diagnosis of EOC, of which ROMA gives the best result; the areas under the curve of these 3 parameters were 0.872, 0.894 and 0.912, respectively. The ROMA index should be widely applied in clinical practice in Vietnam to help in the assessment and management of patients with suspected ovarian cancer.

Ovarian cancer is one of the most common types of cancer of the female reproduction system, occurring in 5–15 per 100,000 women/year in Western countries. According to GLOBOCAN (2012), the incidence in Vietnam is about 3–4.5 per 100,000 women/year, with a prevalence of around 8–10 cases per 100,000 women per year ( Ferlay et al., 2015 ). About 70% of the cases of ovarian cancer are not diagnosed before reaching the advanced stages, and the five-year survival rate associated with ovarian cancer is <30% ( Rauh-Hain et al., 2011 ). Early diagnosis of ovarian cancer is a major factor in improving the survival rate. Markers currently used to distinguish between low-risk and high-risk patients with ovarian cancer include CA-125, Human Epididymis protein 4 (HE4), and the recently introduced Risk of Ovarian Malignancy Algorithm (ROMA) ( Karlsen et al., 2012 ). CA125 is the most widely used biomarker in epithelial ovarian carcinoma (EOC). However, the rather high sensitivity and specificity of CA125 of about 80% in patients in various stages of ovarian cancer drops to 50% or even lower in patients specifically in early stages. Moreover, CA125 level can be elevated in a variety of common benign diseases such as endometriosis and pelvic inflammatory conditions, as well as in borderline tumors of the ovaries ( Montagnana et al., 2011a ). Recently, HE4 has been repeatedly confirmed as one of the most promising biomarkers for early stage diagnosis. HE4 has been found in more than half of ovarian tumors without CA125 expression ( Montagnana et al., 2011b ), and less frequently elevated in benign tumors that mimic the biomarker profile of ovarian cancer, which has often been seen in premenopausal women ( Moore et al., 2012 ). The combination of HE4 and CA125 values resulted in the algorithm to assess the malignancy of the ovaries – ROMA ( Moore et al., 2009 ), which could provide a high sensitivity and specificity for early detection of ovarian cancer. Since the introduction of ROMA into clinical practice, there have been studies of modified cut-off values of these biomarkers and indices ( Winarto et al., 2014 ). Although numerous studies on HE4 and ROMA have been carried out around the world, there are as yet no data from Vietnam. This study aimed to evaluate the validity of serum CA-125, HE4, and ROMA at standard and optimal cut-offs in the preoperative prediction of EOC in Vietnam. Study subjects included 277 women with clinically diagnosed ovarian tumors, including benign ovarian tumors and ovarian cancers, admitted to the Departments of Obstetrics and Gynecology, Hue University Hospital and Hue Central Hospital from January 2016 to November 2017. Selection criteria included patients with sufficient personal information, clinical symptoms, data on serum CA125 and serum HE4 levels, and postoperative pathologic findings. Pregnant women, patients with a history of ovarian, primary peritoneal, or any other associated cancer were excluded from the study. Patients were interviewed for demographic characteristics and history, and then underwent gynecologic examination and pelvic ultrasound to evaluate the tumors, and serum was obtained for the measurements of CA125 and HE4 using Elecsys 2010 system immunoassay ( Elecsys, 2010 ) and ARCHITECT i1000SR system respectively ( ARCHITECT System User Manual, 2009 ). The CA125 cut-off value was 35 U/ml ( Bast et al., 1998 ). The HE4 positive cut-off values for premenopausal and postmenopausal women were > 70 pmol/l and > 140 pmol/l respectively. ROMA was calculated as described elsewhere ( Moore et al., 2010 ). Briefly, a predictive index (PI) was calculated for premenopausal and postmenopausal patients separately using Eqs. (1) , (3) below, followed by insertion of the calculated PI values into Eqs. (2) , (4) , respectively: (1) premenopausal women : PI = − 12.0 + 2 . 38 ∗ LN HE4 + 0.06 26 ∗ LN CA 125 (2) ROMA value % = exp . PI / 1 + exp . PI ∗ 100 ROMA values ≥7.4% or < 7.4% are regarded as high-risk and low-risk respectively. (3) postmenopausal women : PI = − 8.09 + 1 . 04 ∗ LN HE4 + 0.73 2LN CA 125 (4) ROMA = exp . PI / 1 + exp . PI ∗ 100 ROMA values ≥25.3% or < 25.3% are regarded as high-risk and low-risk respectively. Based on the ROMA risk estimates and ultrasound-guided tumor malignancy stratification, appropriate surgical interventions were implemented accordingly. The excised ovarian tumors were examined histologically and classified according to WHO classification. Cases with advanced diseases that were considered inoperable and contain positive ascites were excluded. Preoperative CA125, HE4 and ROMA values were matched postoperatively to the histopathological results to calculate the preoperative prediction values. Data were entered and processed using MedCalc 17.0. Mann-Whitney test was used to compare the two groups having non-standardized distributions. Hanley – McNeil test was used for comparison between two areas under the curves. Ethics approval for study protocol was obtained from the Ethics Committee for Biomedical Researches at Hue University of Medicine and Pharmacy, Hue, Vietnam. Informed consents were obtained from study's subjects.