{"paper_id":"efafd1a5-9d20-49ea-8a22-935b814271e0","body_text":"R E V I E W Open Access\nBiomarkers and algorithms for diagnosis of\novarian cancer: CA125, HE4, RMI and\nROMA, a review\nVincent Dochez 1* , Hélène Caillon 2, Edouard Vaucel 1, Jérôme Dimet 3, Norbert Winer 1 and Guillaume Ducarme 4\nAbstract\nOvarian cancer is the 5th leading cause of death for women with cancer worldwide. In more than 70% of cases, it is\nonly diagnosed at an advanced stage. Our study aims to give an update on the biological markers for\ndiagnosing ovarian cancer, specifically HE4, CA 125, RMI and ROMA algorithms.\nSerum CA125 assay has low sensitivity in the early stag es and can be increased in certain conditions such as\nmenstruation or endometriosis. The level of HE4 is ove rexpressed in ovarian tumors. Its specificity is 94% and\nits level is not affected by endometriosis cysts. The combined measures of CA125 and HE4 have proved to\nbe highly efficient with an area under the curve (AUC) o f up to 0.96. Furthermore, this combined measure of\nC A 1 2 5c a nc o r r e c tt h ev a r i a t i o n si nH E 4w h i c ha r ed u et os m o k i n go rc o n t r a c e p t i o nc o m b i n i n ge s t r o g e np l u s\nprogestin. While the specificity of RMI sometimes reaches 92%, the rather low AUC of 0.86 does not make it\nthe best diagnostic tool. The specificity of ROMA is lower than HE4 (84% compared to 94%).\nTo date, the most efficient biological diagnostic tool to diagnose ovarian cancer is the combination of CA125\nand HE4.\nKeywords: HE4, CA125, ROMA, RMI, Ovarian cancer\nIntroduction\nOvarian cancer is the 5th worldwide leading cause of\ndeath of women due to cancer [ 1]. In more than 70% of\ncases, it is diagnosed at an advanced phase. The progno-\nsis for ovarian cancer remains poor overall, with a 46%\n5-year survival rate [ 2]. The prognosis is closely related\nto the stage at diagnosis: survival rate of > 70% after 5\nyears for stage I or II, survival rates between 20 and 40%\nfor stage III or IV [ 3, 4].\nMore than 90% of benign tumors are found in pre-\nmenopausal patients who have been operated on,\nwhereas in postmenopausal patients only 60% of tumors\nare benign [ 5]. It seems essential to differentiate early\nmalignant ovarian tumors to benign ovarian tumors.\nExams are therefore needed which must be prioritized in\norder to advise the patient (monitoring, treatment or\nsurvey) according to the lesion, and first and foremost to\nthe clinical history of the patient [ 6].\nConcerning laboratory exams, several tumor biomarkers\nhave been evaluated. The Carbohydrate Antigen 125\n(CA125) was first described in the early 1980 ’s[ 7]. In cases\nof ovarian cancer, serum CA125 level may be elevated, but\nthis marker has a low sensitivity in the early stages of\novarian cancer [ 8]. Increased CA125 levels are also re-\nported in other physiological or pathological conditions,\nsuch as menstruation, pregnancy, endometriosis and in-\nflammatory diseases of the peritoneum [ 9]. Other bio-\nmarkers have been developed in order to improve\nspecificity for ovarian carcinomas, such as the Human\nEpididymis Protein 4 (HE4) [ 10]. This biomarker is re-\nported to be overexpressed in ovarian cancer [ 11].\nAlthough the specificity of these markers is rather reliable,\nthey are not very sensitive. For this reason, algorithms,\nRMI (Risk of Malignancy Index) and ROMA (Risk of\nOvarian Malignancy Algorithm), were developed in an at-\ntempt to improve the inherent characteristics of these\nbiomarkers.\nHE4 was found to be a reliable biological marker for de-\ntecting ovarian cancer (level of evidence [LE]1) and ROMA\n© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0\nInternational License ( http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and\nreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to\nthe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver\n(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.\n* Correspondence: vincent.dochez@chu-nantes.fr\n1Service de Gynécologie-Obstétrique, CHU de Nantes, 44093 Nantes, France\nFull list of author information is available at the end of the article\nDochez et al. Journal of Ovarian Research           (2019) 12:28 \nhttps://doi.org/10.1186/s13048-019-0503-7\n\nalgorithm was more sensitive but less specific than HE4\nalone (LE2) for the French National College of Obstetri-\ncians and Gynaecologists (CNGOF) [12]. CNGOF also con-\ncluded that complementary studies should be necessary\nbefore using HE4 on a routine basis (Grade B). Our study\nintends to provide a comprehensive update for ovarian can-\ncer diagnosis using biomarkers (HE4 and CA125) and algo-\nrithms (RMI and ROMA). The sensitivity, specificity,\npositive and negative predict i v ev a l u e so ft h e s et o o l sh a v e\nbeen collated in Table 1 for CA125 and HE4 and T able 2\nfor RMI and ROMA algorithms. Only the values of some\ncited articles are referenced in order to lighten the table\nand make it readable.\nDiscussion\nCarbohydrate antigen 125 (CA 125)\nCarbohydrate Antigen 125 (CA 125), sometimes named\nas Cancer Antigen 125 or Tumor Antigen 125, is a\nmucin-type glycoprotein, produced by the MUC16 gene,\nand associated with the cellular membrane.\nThis biomarker is most often used for ovarian lesions.\nIt has been used in the early 1980 ’s when Bast et al. [ 7]\nspecifically isolated the monoclonal antibody OC125 in\ncancerous ovarian tissue compared to healthy ovarian\ntissue. Its upper limit is 35 U/mL in pre and\npost-menopausal patients [ 13]. However, this measure-\nment is not very sensitive in the early phases of ovarian\ncancer (only reported to be elevated in 23 to 50% of\nstage I cases) [ 10]. In addition, elevated serum CA125\nlevels may be observed in other physiological or patho-\nlogical conditions (menstruation, pregnancy, endometri-\nosis, inflammatory diseases of the peritoneum) [ 9]. In a\nmeta-analysis by Ferraro et al. [ 14], the specificity of\nCA125 for detecting ovarian cancer was 78% (95%CI 76 –\n80). To describe tumor markers and screening tests, the\nReceiver Operating Characteristic (ROC) area under the\ncurve (AUC) is frequently employed since it represents a\nuseful graphic tool for comparing biomarkers and algo-\nrithms. The ROC measures the discrimination of a test,\ni.e. its ability to distinguish between having disease and\nnot having disease for a given patient. In the study by\nDikmen et al. [ 15] ,t h eA U Cf o rC A 1 2 5w a sr a t h e rw e a k\n(0.78), suggesting that it was probably not the ideal\nmarker for diagnosing ovarian cancer.\nSerum CA125 levels were frequently measured\nwhen ovarian cysts are observed, in order to rule out\na malignant tumor. But for the past several decades,\nelevated serum CA125 levels have been seen in endo-\nmetrioma, thus giving a high rate a false positives\n[16]. This was confirmed in a recent Cochrane review\nwhich reported that among the 97 biomarkers stud-\nied, CA125 was the only marker which is elevated in\ncases of endometrioma with 40% sensitivity and 91%\nspecificity with a cut off limit of 35 U/mL [ 17]. In\nanother very recent meta-analysis, Hirsch et al. [ 18]\ndemonstrated that CA125 should be useful to diag-\nnose endometriosis, especially with an increasing sen-\nsitivity corresponding simultaneously with the disease\nphase. Chen et al. [ 19]r e p o r t e dt h a tC A 1 2 5l e v e l s\nwere significantly higher in the group with endome-\ntriotic cysts compared to group with other benign\novarian tumors (49.7 U/mL vs. 21.6 U/mL).\nSince CA125 has been a tumor marker for several de-\ncades, the changes in its levels according to the patient ’s\ndemographic characteristics and lifestyles have been\nevaluated numerous times [ 20, 21]. Smoking does not\nappear to modify serum CA125 levels [ 22, 23]. Whereas\nCA125 may vary during menstrual cycles, it has been\ndemonstrated that levels are not affected by a contracep-\ntion combining estrogen plus progestin [ 21, 24]. Likewise,\nthe body mass index does not appear to modify CA125\nlevels [25].\nIn practice, CA125 is often measured in cases of ovarian\ncysts, but according to its low specificity and the observed\nincreased levels in different physiological situations, it is\nnot considered as a very good differentiating biomarker\nfor ovarian tumors. For this reason, new biomarkers have\nbeen evaluated in an attempt to improve early diagnosis of\novarian cancer [26].\nHuman epididymis protein 4 (HE4)\nHuman Epididymis Protein 4 (HE4) is a new biomarker\nwhich has been currently evaluated for diagnosing ovar-\nian malignant tumors [ 10]. It is a glycoprotein belonging\nto the family of whey acidic four-disulfide core proteins,\naccounting for its alternative name of WFDC2 and the\nlarger protein family called “WAP” for whey acidic pro-\nteins. The main genes coding for the WAP proteins are\nmainly located on chromosome 20q12 –13.1 [ 27].\nPresent in whey, these proteins are called WAP , which is\ncomposed of around 50 amino acids, and its biological\nfunction has not yet been completely identified [ 28].\nHE4, which contains 2 WAP domains, was initially\nisolated in the epididymis and might play a role in sperm\nmaturation [ 29]. This biomarker is weakly expressed in\nthe epithelium tissues of respiratory and reproductive\norgans, but is overexpressed in ovarian tumors, espe-\ncially in endometrioid ovarian cancer [ 11]. In addition, it\nappears that HE4 is not as strongly expressed in clear\ncell ovarian carcinomas as in other epithelial ovarian\ncancers [ 30]. Yanaranop et al. [ 31] reported a specificity\nof 86% for HE4, and the AUC was higher than CA125\nalone, with values of 0.893 and 0.865, respectively [ 32].\nThese data, in accordance with those reported in a re-\ncent Italian multicentre study included 387 patients,\nshowed that HE4 for diagnosing ovarian epithelial can-\ncer appeared more reliable than CA125 [ 33].\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 2 of 9\n\nTable 1 Diagnostic performance of CA125, HE4 and combination of CA125 + HE4 in the subset of studies cited in this article\nSystematic\nreview or\nmeta-\nanalysis\nCA125 HE4 CA125 + HE4\nSe (%)\n(95% IC)\nSp (%)\n(95% IC)\nPPV\n(%)\nNPV\n(%)\nAUC (95% IC) Se (%)\n(95% IC)\nSp (%)\n(95% IC)\nPPV\n(%)\nNPV\n(%)\nAUC (95% IC) Se (%)\n(95% IC)\nSp (%)\n(95% IC)\nPPV\n(%)\nNPV\n(%)\nAUC (95% IC)\nFerraro\net al. [ 14]\nX 79 (77 –82) 78 (76 –80) 79 (76 –81) 93 (92 –94) 82 (78 –86) 76 (72 –80)\nDikmen\net al. [ 15]\n63 0.78 78 0.93\nChen et\nal. [ 19]\n93 67 0.93 (0.88 –.97) 73 99 0.96 (0.93 –1) 97 66 0.96 (0.93 –1)\nYanaranop\net al. [ 31]\n84 53 41 89 0.81 (0.74 –0.87) 66 86 65 87 0.82 (0.76 –.89)\nWilailak\net al. [ 32]\n0.87 (0.80–0.93) 0.89 (0.84 –0.95) 0.89 (0.84 –0.95)\nWang\net al. [ 36]\nX 79 (74 –84) 82 (77 –87) 0.87 (0.84 –0.90) 76 (72 –80) 94 (90 –96) 0.89 (0.86 –0.92)\nZhen et\nal. [ 37]\nX 74 (72 –76) 83 (81 –84) 0.85 74 (72 –76) 90 (89 –91) 0.89\nAbdel-Azeez et\nal. [ 45]\n73 (57–86) 79 (58 –93) 0.90 (0.82 –0.97) 83 (68 –93) 88 (68 –97) 0.95 (0.90 –1) 90 79\nHolcomb\net al. [ 46]\n85 (69–95) 59 (52 –66) 27 96 65 (46 –80) 92 (87 –95) 58 94 91 55 26 97\nMoore et al.\n[48]\n61 0.84 (0.77 –0.90) 78 0.91 (0.86 –0.95) 81 0.91 (0.87 –0.96)\nGoff et al.\n[52]\n79 (67–88) 76 (68 –83) 63 87 58 (45 –70) 94 (88 –97) 83 81\nMeys et al.\n[55]\nX\nVan Gorp\net al. [ 56]\n80 (72–85) 82 (76 –86) 75 (67 –81) 83 (78 –88)\nAl Musalhi\net al. [ 57]\n79 62 38 91 0.81 71 90 68 91 0.82\nMoore et al.\n[62]\n89 (85–93) 75 (70 –79) 60 94\nLi et al. [ 64] X 77 (58 –89) 84 (76 –90) 0.88 (0.85 –0.91) 79 (74 –84) 93 (87 –96) 0.82 (0.78 –0.85)\nWei et al. [ 66] 8 59 2 9 1 8 9 7 59 8 9 6 8 5\nSandri et al.\n[67]\n91 71 0.90 (0.86 –0.93) 83 91 0.92 (0.89 –.95)\nCA125 carbohydrate antigen 125, HE4 human epididymis protein 4, Se sensitivity, Sp specificity, PPV positive predictive value, NPV negative predictive value, AUC area under the curve\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 3 of 9\n\nTo measure this biomarker, an immune-enzymatic\n(EIA) assay described by the Fujirebio lab was developed\n[34]. Other immunological methods have since been de-\nveloped and evaluated such as electrochemiluminescent\n(ECLIA) or chemiluminescent microparticle immuno-\nassay (CMIA). A recent study evaluated these different\nmeasurement techniques for HE4 levels [ 35]. The au-\nthors reported that ECLIA and CMIA were well com-\npared with the reference method and could be routinely\nused in practice. However, significantly different mean\nvalues were reported according to the immunological\nassay used; HE4 marker levels might always be interpreted\naccording to the immunological assay employed. The cut\noff level of 70 pmol/L is often used for pre-menopause pa-\ntients and 140 pmol/L for menopause patients; but some-\ntimes the threshold level of 140 pmol/L is employed, or\neven other close but outlying values. The threshold value\nof HE4 used, including whether or not the menopausal\nstatus, is left to the choice of clinicians [ 36, 37]. Indeed,\nthese values can be used whatever the immunological\nmethod performed. Nevertheless, given the fact that is\nproven that HE4 increases significantly over the age, the\nuse of 2 thresholds (70 and 140 pmol/l) seems preferable\nthan the use of a single threshold (140 pmol/l) [ 38, 39]. In\nconclusion, the review of the literature demonstrates that\nall meta-analyses always take into account the different\nmethods used to measure HE4 and the results must be\ncarefully handled.\nAlthough CA125 levels seem to be elevated in endo-\nmetrioma, HE4 levels appear to remain stable [ 19, 40,\n41]. HE4 levels in patients with endometrioma were\ncomparable to levels in patients with other benign ovar-\nian cysts (53.0 pmol/L vs. 52.8 pmol/L) [ 19], quite under-\nstandable since the gene coding for HE4 is not\noverexpressed in endometriotic lesions [ 42]. A recent\nstudy has also confirmed that serum HE4 was a better\nTable 2 Diagnostic performance of RMI and ROMA algorithms in the subset of studies cited in this article\nSystematic\nreview or\nmeta-\nanalysis\nRMI ROMA\nSe (%)\n(95% IC)\nSp (%)\n(95% IC)\nPPV\n(%)\nNPV\n(%)\nAUC (95% IC) Se (%)\n(95% IC)\nSp (%)\n(95% IC)\nPPV\n(%)\nNPV\n(%)\nAUC (95% IC)\nFerraro\net al. [ 14]\nX\nDikmen\net al. [ 15]\n88 0.96\nChen et al. [ 19] 97 80 0.97 (0.95 –1)\nYanaranop et al.\n[31]\n78 80 60 90 0.88 (0.83 –0.93) 84 69 52 91 0.86 (0.81 –0.91)\nWilailak et al.\n[32]\n0.84 (0.77–0.91) 0.86 (0.81 –0.91)\nWang et al.\n[36]\nX 85 (81 –89) 82 (77 –87) 0.91 (0.88 –0.93)\nZhen et al.\n[37]\nX\nAbdel-Azeez et\nal. [ 45]\nHolcomb et al.\n[46]\nMoore et al. [ 48]\nGoff et al. [ 52]\nMeys et al. [ 55] X 75 (72 –79) 92 (88 –94)\nVan Gorp et al.\n[56]\nAl Musalhi et al.\n[57]\n77 82 56 93 0.85 75 88 65 92 0.84\nMoore et al. [ 62]\nLi et al. [ 64] X 89 (84 –93) 83 (77 –88) 0.93 (0.90 –0.95)\nWei et al. [ 66]9 4 9 3 9 0 8 6\nSandri et al. [ 67] 89 81 0.93 (0.90 –0.96)\nRMI risk of malignancy index, ROMA risk of ovarian malignancy algorithm, Se sensitivity, Sp specificity, PPV positive predictive value, NPV negative predictive value,\nAUC area under the curve\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 4 of 9\n\ndiagnostic biomarker than CA125 in ovarian cancer pa-\ntients with endometriosis [ 38].\nVariations of HE4 levels in variable situations were\nalso evaluated. Whereas Bolstad et al. [ 38] reported\nmodified levels of HE4 according to Body Mass Index\n(BMI), Ferraro et al. [ 25] did not report significant dif-\nferent levels of HE4 in 103 patients according to BMI,\nprobably explained by inclusion of men and women in\nthis study which should constitute a bias. In conclusion,\nserum HE4 levels appeared to not be modified by the\nBMI, as CA125.\nNevertheless, in opposite to CA125, smoking seems to\nbe a significant factor which affects serum HE4 varia-\ntions [ 43]. HE4 level is increased from 20 to 30% in\nsmokers compared to non-smokers [ 23, 38, 44]. So, HE4\nlevel should be always interpreted carefully in smokers,\nsince it could be misinterpreted as a false positive result.\nIn opposite to CA125, contraceptive use contributes to\nvariations in HE4 levels. Ferraro et al. [ 24] reported a\nsignificantly lower level of HE4 in patients using oral\ncontraception compared to patients using other contra-\nceptive methods ( p = 0.008). Therefore, in order to mis-\ninterpret HE4 levels, it appears noteworthy to include\nthe contraceptive method in the patient ’s clinical history.\nAssociating HE4 + CA125\nThree meta-analyzes or systematic reviews show the use\nof HE4 or CA125, always with different thresholds for\nHE4 [ 14, 36, 37]. The combined use of these markers is\nonly sometimes studied [ 45–47]. Chen et al. [ 19] re-\nported a specificity of 65.7% using the ECLIA immuno-\nlogical method with a cut-off value for HE4 of 140\npmol/L. In a different study using another technique to\nassess serum HE4, the specificity of the association\nCA125 and HE4 was much better (80%) [ 48]. However,\nthe ROC AUC when combining the two markers was\nhigh, varying from 0.96 (IC95% 0.93 –1) [ 19] to 0.91\n(IC95% 86.7 –96.0) [ 48]. Thus, the association of CA125\nand HE4 is a useful diagnostic tool in ovarian cancer\nand may be used in addition to each biomarker.\nThese results are confirmed by another study ana-\nlyzing each marker separately or in combination [ 49].\nThat is, in case of increased HE4 and CA125, the\nspecificity was better. The use of a third marker, for\nexample clinical, could also improve the detection of\nthese ovarian cancers.\nBesides its use diagnosing ovarian cancer, the measure-\nments of the association HE4 + CA125 can also be useful\nin differential diagnosis of different ovarian tumors.\nAnastasi et al. [ 50] studied 57 patients with endome-\ntrioma; all had increased CA125 levels, while the HE4\nlevels remained normal. Another very recent study con-\nfirmed the significant difference between HE4 and CA125\nlevels in endometrioma cases [ 51]. Thus, endometrioma\ncould be suspected whenever CA125 levels are elevated\nwhile HE4 levels remain normal. In addition, while HE4\nlevel varies in smokers and in contraceptive combining es-\ntrogen plus progestin users, simultaneous CA125 evalu-\nation which is not affected by these variables should allow\nbetter interpretation of abnormal HE4 levels.\nIn conclusion, it seems worthwhile to measure both\nmarkers in cases of suspected benign ovarian tumors: an\nincreased value of the 2 markers being suggestive of an\novarian cancer As suggested in a recent study by Goff et\nal., thresholds of 70 and 140 pmol/l according to meno-\npausal status and 35UI/ml for CA125 seems preferable\n[52]. Furthermore, the use of this combined HE4 and\nCA125 assay may also be of major interest in ovarian\ncancer screening in the general population, as shown by\nUrban et al. [ 44]. However, in this context, the place of\ntransvaginal ultrasound remains to be clarified, and the\nuse of a second positive test could reduce sensitivity (key\nprinciple of a screening test).\nRisk of malignancy index (RMI)\nRMI was proposed in 1990 by Jacobs et al. [ 53], using\nCA125, ultrasound findings and menopausal status ac-\ncording to the formula: RMI = U x M x CA125 with U =\nultrasound score (U = 0 if ultrasound score = 0, U = 1 if\nultrasound score = 1, U = 3 if ultrasound score 2 to5), M\n= menopause status (M = 1 for pre-menopausal women,\nM = 3 for post-menopausal women). A RMI score above\n200 proved to have a strong association with a high risk\nof malignancy (sensitivity 85.4% and specificity 96.9%).\nAnother study performed in 2012 on nearly 1000 pa-\ntients demonstrated that the ultrasound assessment was\nsuperior according to IOTA (International Ovarian\nTumor Analysis) criteria compared to RMI [ 54]. These\ndata were confirmed in a recent meta-analysis published\nin 2016 [ 55]. This study of almost 20,000 ovarian tumors\nreported better results with the use of ultrasound criteria\n(sensitivity 93% and specificity 80%) compared to that\nobtained with the RMI algorithm (sensitivity 75% and spe-\ncificity 92%). The specificity of RMI for diagnosing ovarian\ncancer is rather high, 92.4% [ 56] and 92% in a recent\nmeta-analysis [55]. This specificity could be increased by\nmodifying the threshold level for malignancy (using 250,\nas reported by Al Musalhi et al. [ 57]).\nSince 1990 with Jacobs ’ suggestion to use RMI algo-\nrithms, several variations in RMI formula have been de-\nveloped. Chopra et al. [ 58] studied 100 patients, using a\nmodified RMI with the maximum values for the U et M\nparameters were 4 instead of 3, which increased the sen-\nsitivity to 96.7% with 84% specificity, and a positive pre-\ndictive value of 85.5%. Recently, the use of four different\nRMI formulas showed the same results for sensitivity.\nThe positive likelihood ratio was reported between 3.52\nand 4.41 [ 59].\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 5 of 9\n\nFor screening tests, the use of ROC AUC seems to be\na useful indicator for detecting cancer. That of RMI re-\nmains weak (0.86) [ 60] compared to other biomarkers\nwhich emphasize that RMI algorithm don ’t appear to be\nthe most useful diagnostic tool for ovarian cancer.\nAccording to this hypothesis, a 2-steps triage model,\nassociating ultrasound findings first, RMI could improve\nscreening results. Using this 2-step model, a recent study\nin 2016 reported an improved detection rate of ovarian\ncancer through 72 to 85% [ 61].\nFinally, concerning this RMI algorithm or modified RMI\nalgorithms, the variable U may be equal to 0, the RMI\nscore may be 0 (RMI = U x M x CA125). This value 0 may\nsometimes seem aberrant in the interpretation of a score.\nNevertheless, the RMI score is based on a threshold (200).\nThe calculation of specificity, sensitivity and positive and\nnegative predictive values is not based on the value of the\nscore but on its value below or above that threshold. The\nvalue 0 therefore impacts the interpretation of this algo-\nrithm only very moderately.\nRisk of ovarian malignancy algorithm (ROMA)\nIn 2009, Moore proposed a new algorithm: Risk of Ovar-\nian Malignancy Algorithm (ROMA) [ 62]. He associated\nHE4 and CA125 levels according to the menopausal sta-\ntus, defined by lack of menstruation or clinical signs of\nmenopause for 6 months.\nPre−menopausal Predictive Index PIðÞ\n¼ −12:0 þ 2:38 /C2 LN HE4ðÞ þ 0:0626 /C2 LN CA125ðÞ\nPost−menopausal Predictive Index PIðÞ\n¼ −8:09 þ 1:04 /C2 LN HE4ðÞ þ 0:732 /C2 LN CA125ðÞ\nPredicted Probability PPðÞ\n¼ exp PIðÞ = 1 þ exp PIðÞ½/C138 /C2 100\nTherefore, the ROMA score corresponds to Predicted\nProbability [PP] and is expressed by a percentage rate.\nDifferent cut off levels are proposed for non-menopausal\nwomen, and another for women having reached meno-\npause. According to the immunological assay for meas-\nuring CA125 and HE4, the cut off levels can differ to\nclassify patients into either a low or an high risk group\n[63]. In fact, with the Roche Diagnostics Laboratory ’s\nECLIA method, the cut off level to classify patients in a\nhigh risk group was 11.4% for pre-menopausal patients,\nand 29.9% for menopause patients. Whereas with Abbott\nDiagnostics Laboratory ’s CMIA method, the cut off\nlevels were respectively 7.4 and 25.3% [ 32]. It is thus in-\ndispensable to know which method is used or to\nrefer to the normal values given by the lab perform-\ning the tests in order to correctly interpret the re-\nsults of this algorithm.\nIn a meta-analysis, the ROMA algorithm was reported\nwith less specificity than that associated with HE4 levels\n(84% vs. 94%), but a better correlation than with CA125\nlevels (84% vs. 78%) [ 64]. In this meta-analysis, the\nAUC for the ROMA algorithm was better than HE4\nor CA125 (0.93, 0.82 and 0.88; respectively), as in the\nmeta-analysis performed by Wang et al. (0.91, 0.89\nand 0.87; respectively) [ 36].\nAnother meta-analysis by Kaijser et al. [ 65] demon-\nstrated a ROMA sensitivity between 76 and 86%, while\nspecificity was evaluated between 74 and 95%, in spite of\nusing different methods for measuring the markers.\nRecently, ROMA was reported having a non-significant\ndifferent specificity to CA125 (approximately 92.5%) but\nwith significantly improved sensitivity levels: 93.7% vs.\n85.0% [ 66]. For Chen et al. [ 19], ROMA was more sensi-\ntive than HE4, respectively 96.7 and 73.3%, but with less\nspecificity (80% vs. 98.6%). In this same study, AUC for\nROMA and HE4 were not significantly different (0.97\nand 0.96, respectively). Several more recent studies con-\nfirm these results [ 32, 67]. Nevertheless, it should be\nnoted that it is difficult to interpret these meta-analyses\nresults. When studies have used different measuring tech-\nniques for CA125 and HE4 [ 14, 36, 64], the results thus\nindirectly implied different ROMA calculations. There-\nfore, one must be cautious in interpreting cut off levels.\nAlthough the dual measurement of CA125 and HE4 is\napparently the best diagnostic tool over and above\nROMA algorithms; the fact of including ages in the\nROMA model could be a valuable contribution for diag-\nnosing ovarian cancer [ 68]. It was reported that serum\nHE4 levels regularly increased with age, without any\nsudden peak at menopause [ 38, 39]. Using age in the\nalgorithm, it should be a valuable contribution for evalu-\nating serum HE4 levels.\nModified risk of ovarian malignancy algorithm: CPH-I and\nROMA P\nIn 2015, the Copenhagen Index (CPH-I) was reported as\na novel diagnostic score index in ovarian tumors [ 69]. It\nused the same mathematical method than the ROMA\nalgorithm with a Predicted Probability called PP. The\nCPH-I formula is: CPH-I =− 14.0647 + 1.0649 x log2(HE4) +\n0.6050 x log2(CA125) + 0.2672 x age / 10 with PP = e(CPH-I)\n/( 1+e ( C P H - I ) ) .\nThe ROC AUC were comparable according to the dif-\nferent respective algorithms: CPH-I, ROMA and RMI\n(0.96, 0.95 and 0.96, respectively). Thus, the introduction\nof the age in the algorithm did not improve the diagnostic\nof ovarian cancer. More recently, Chudecka-Glaz et al.\n[70] evaluated another modified ROMA algorithm, called\nROMA P , which took into account the patient ’s age and\nnot her menopausal status, according to the formula:\nROMA P = exp.(PI) / (1-exp(PI)) × 100 with the predictive\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 6 of 9\n\nindex formula PI: PI = A + W(HE4) x ln(HE4) +\nW(CA125) x ln(CA125), and A, W(HE4) et W(CA125)\nwere the varying coefficients for each decade in function\nof the patient ’s age. ROMA P algorithm had a higher spe-\ncificity and a higher positive predictive value, but the sen-\nsitivity and the negative predictive value were lower than\nthe non-modified ROMA algorithm. Otherwise, the ROC\nAUC for these two algorithms were comparable (0.923 for\nthe ROMA P vs. 0.934 for ROMA).\nConclusions\nThe best biological diagnostic tool today seems to be a\ncombination of CA125 and HE4 levels in order to predict\nthe risk of ovarian cancer in patients with suspected be-\nnign ovarian tumors. If the level of CA125 is increased as\nwell as that of HE4, it is necessary to evoke a malignant le-\nsion and therefore to envisage a surgical treatment for an\nanatomopathological examination. On the other hand, if\none of the markers was above the cut-off as long as the\nother was below the cut-off specified, a simple ultrasound\nor biological monitoring may be considered. As the HE4\nlevels increase with advancing age, it might be interesting\nto establish algorithms which take into account the pa-\ntients’ age and not her menopausal status. The previously\npublished algorithms (CHP-I or ROMA P) have not\nproved to be valuable compared to RMI or ROMA algo-\nrithms. Serum HE4 levels vary in smokers and in hormo-\nnal contraceptive users, thus it seems relevant that this\ninformation should always be included in the patient ’s\nclinical history. Nonetheless, since CA125 levels are inde-\npendent from these variables, the simultaneous measure\nof these two markers allows the correction of any possible\nvariations in such specific cases.\nAbbreviations\nAUC: Area under the Curve; BMI: Body Mass Index; CA125: Carbohydrate\nAntigen 125; CMIA: Chemiluminesc ent microparticle immunoassay;\nECLIA: Electrochemiluminescent immunoassay; EIA: Immune-Enzymatic\nassay; HE4: Human Epididymis Protein 4; IOTA: International Ovarian Tumor\nAnalysis; RMI: Risk of Malignancy Index; ROC: Receiver Operating Characteristic;\nROMA: Risk of Ovarian Malignancy Algorithm\nAcknowledgements\nNot applicable\nFunding\nNot applicable\nAvailability of data and materials\nNot applicable\nAuthors’ contributions\nVD and GD developed the original design. VD performed the initial literature\nreview. VD and GD wrote the first draft of the report. EV, JD, HC and NW\nparticipated in the development and the revisions of the manuscript. All\nauthors contributed to the writing of the final manuscript. All authors read\nand approved the final manuscript.\nEthics approval and consent to participate\nNot applicable\nConsent for publication\nNot applicable\nCompeting interests\nThe authors declare that they have no competing interest.\nPublisher’sN o t e\nSpringer Nature remains neutral with regard to jurisdictional claims in\npublished maps and institutional affiliations.\nAuthor details\n1Service de Gynécologie-Obstétrique, CHU de Nantes, 44093 Nantes, France.\n2Service de Biochimie, CHU de Nantes, Nantes, France. 3Centre de recherche\nclinique, Centre Hospitalier Départemental Vendée, La Roche sur Yon, France.\n4Service de Gynécologie-Obstétrique, Centre Hospitalier Départemental\nVendée, La Roche sur Yon, France.\nReceived: 8 May 2018 Accepted: 19 March 2019\nReferences\n1. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2008. CA Cancer J Clin.\n2008;58:71–96.\n2. Chan JK, Teoh D, Hu JM, et al. Do clear cell ovarian carcinomas have poorer\nprognosis compared to other epithelial cell types? A study of 1411 clear cell\novarian cancers. Gynecol Oncol. 2008;109:370 –6.\n3. Heintz APM, Odicino F, Maisonneuve P, et al. Carcinoma of the ovary. FIGO\n26th annual report on the results of treatment in gynecological Cancer. Int\nJ Gynaecol Obstet. 2006;95(Suppl 1):S161 –92.\n4. Zeppernick F, Meinhold-Heerlein I. The new FIGO staging system for\novarian, fallopian tube, and primary peritoneal cancer. Arch Gynecol Obstet.\n2014;290:839–42.\n5. Enakpene CA, Omigbodun AO, Goecke TW, et al. Preoperative evaluation\nand triage of women with suspicious adnexal masses using risk of\nmalignancy index. J Obstet Gynaecol Res. 2009;35:131 –8.\n6. Brun J-L, Fritel X, Levêque J. Recommandations pour la pratique clinique :\ntumeurs ovariennes présumées bénignes – Objectifs, méthodes et\norganisation. J Gynecol Obstet Biol Reprod. 2013;42:710 –4.\n7. Bast RC, Feeney M, Lazarus H, et al. Reactivity of a monoclonal antibody\nwith human ovarian carcinoma. J Clin Invest. 1981;68:1331 –7.\n8. Urban N, McIntosh MW, Andersen M, Karlan BY. Ovarian cancer screening.\nHematol Oncol Clin North Am. 2003;17:989 –1005 ix.\n9. Buamah P. Benign conditions associated with raised serum CA-125\nconcentration. J Surg Oncol. 2000;75:264 –5.\n10. Hellström I, Raycraft J, Hayden-Ledbetter M, et al. The HE4 (WFDC2) protein\nis a biomarker for ovarian carcinoma. Cancer Res. 2003;63:3695 –700.\n11. Drapkin R, von Horsten HH, Lin Y, et al. Human epididymis protein 4 (HE4)\nis a secreted glycoprotein that is overexpressed by serous and\nendometrioid ovarian carcinomas. Cancer Res. 2005;65:2162 –9.\n12. Lahlou N, Brun J-L. Marqueurs sériques et tumoraux ovariens dans le\ndiagnostic des tumeurs ovariennes présumées bénignes. J Gynecol Obstet\nBiol Reprod. 2013;42:752 –9.\n13. Bast RC, Klug TL, St John E, et al. A radioimmunoassay using a monoclonal\nantibody to monitor the course of epithelial ovarian cancer. N Engl J Med.\n1983;309:883–7.\n14. Ferraro S, Braga F, Lanzoni M, et al. Serum human epididymis protein 4 vs\ncarbohydrate antigen 125 for ovarian cancer diagnosis: a systematic review.\nJ Clin Pathol. 2013;66:273 –81.\n15. Dikmen ZG, Colak A, Dogan P, et al. Diagnostic performances of CA125, HE4,\nand ROMA index in ovarian cancer. Eur J Gynaecol Oncol. 2015;36:457–62.\n16. Markman. The role of CA-125 in the Management of Ovarian Cancer.\nOncologist. 1997;2:6–9.\n17. Nisenblat V, Bossuyt PMM, Shaikh R, et al. Blood biomarkers for the non-\ninvasive diagnosis of endometriosis. Cochrane Database Syst Rev. 2016:\nCD012179.\n18. Hirsch M, Duffy J, Davis CJ, et al. Diagnostic accuracy of cancer antigen 125\nfor endometriosis: a systematic review and meta-analysis. BJOG. 2016;123:\n1761–8.\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 7 of 9\n\n19. Chen X, Zhou H, Chen R, et al. Development of a multimarker assay for\ndifferential diagnosis of benign and malignant pelvic masses. Clin Chim\nActa. 2014;440C:57–63.\n20. Kafali H, Artunc H, Erdem M. Evaluation of factors that may be responsible\nfor cyclic change of CA125 levels during menstrual cycle. Arch Gynecol\nObstet. 2007;275:175 –7.\n21. Lowe KA, Shah C, Wallace E, et al. Effects of personal characteristics on\nserum CA125, mesothelin, and HE4 levels in healthy postmenopausal\nwomen at high-risk for ovarian cancer. Cancer Epidemiol Biomark Prev.\n2008;17:2480–7.\n22. Dehaghani AS, Ghiam AF, Hosseini M, et al. Factors influencing serum\nconcentration of CA125 and CA15-3 in Iranian healthy postmenopausal\nwomen. Pathol Oncol Res. 2007;13:360 –4.\n23. Urban N, Thorpe J, Karlan BY, et al. Interpretation of single and serial\nmeasures of HE4 and CA125 in asymptomatic women at high risk for\novarian cancer. Cancer Epidemiol Biomark Prev. 2012;21:2087 –94.\n24. Ferraro S, Schiumarini D, Panteghini M. Human epididymis protein 4: factors\nof variation. Clin Chim Acta. 2015;438:171 –7.\n25. Ferraro S, Borille S, Caruso S, et al. Body mass index does not influence human\nepididymis protein 4 concentrations in serum. Clin Chim Acta. 2015;446:163–4.\n26. Yang W-L, Lu Z, Bast RC. The role of biomarkers in the management of\nepithelial ovarian cancer. Expert Rev Mol Diagn. 2017;17:577 –91.\n27. Bouchard D, Morisset D, Bourbonnais Y, Tremblay GM. Proteins with whey-\nacidic-protein motifs and cancer. Lancet Oncol. 2006;7:167 –74.\n28. Hennighausen LG, Sippel AE. Mouse whey acidic protein is a novel member\nof the family of “four-disulfide core ” proteins. Nucleic Acids Res. 1982;10:\n2677–84.\n29. Kirchhoff C, Habben I, Ivell R, Krull N. A major human epididymis-specific\ncDNA encodes a protein with sequence homology to extracellular\nproteinase inhibitors. Biol Reprod. 1991;45:350 –7.\n30. Shridhar V, Lee J, Pandita A, et al. Genetic analysis of early- versus late-stage\novarian tumors. Cancer Res. 2001;61:5895 –904.\n31. Yanaranop M, Anakrat V, Siricharoenthai S, et al. Is the risk of ovarian\nmalignancy algorithm better than other tests for predicting ovarian\nmalignancy in women with pelvic masses? Gynecol Obstet Investig. 2017;\n82:47–53.\n32. Wilailak S, Chan KK, Chen CA, et al. Distinguishing benign from malignant\npelvic mass utilizing an algorithm with HE4, menopausal status, and\nultrasound findings. J Gynecol Oncol. 2015;26:46 –53.\n33. Romagnolo C, Leon AE, Fabricio ASC, et al. HE4, CA125 and risk of ovarian\nmalignancy algorithm (ROMA) as diagnostic tools for ovarian cancer in\npatients with a pelvic mass: an Italian multicenter study. Gynecol Oncol.\n2016;141:303–11.\n34. Ferraro S, Panteghini M. Is serum human epididymis protein 4 ready for\nprime time? Ann Clin Biochem. 2014;51:128 –36.\n35. Ferraro S, Borille S, Carnevale A, et al. Verification of the harmonization of\nhuman epididymis protein 4 assays. Clin Chem Lab Med. 2016;54:1635 –43.\n36. Wang J, Gao J, Yao H, et al. Diagnostic accuracy of serum HE4, CA125 and\nROMA in patients with ovarian cancer: a meta-analysis. Tumour Biol. 2014;\n35:6127–38.\n37. Zhen S, Bian L-H, Chang L-L, Gao X. Comparison of serum human\nepididymis protein 4 and carbohydrate antigen 125 as markers in ovarian\ncancer: a meta-analysis. Mol Clin Oncol. 2014;2:559 –66.\n38. Bolstad N, Øijordsbakken M, Nustad K, Bjerner J. Human epididymis protein\n4 reference limits and natural variation in a Nordic reference population.\nTumour Biol. 2012;33:141 –8.\n39. Moore RG, Miller MC, Eklund EE, et al. Serum levels of the ovarian cancer\nbiomarker HE4 are decreased in pregnancy and increase with age. Am J\nObstet Gynecol. 2012;206:349.e1 –7.\n40. Terry KL, Sluss PM, Skates SJ, et al. Blood and urine markers for ovarian\ncancer: a comprehensive review. Dis Markers. 2004;20:53 –70.\n41. Huhtinen K, Suvitie P, Hiissa J, et al. Serum HE4 concentration differentiates\nmalignant ovarian tumours from ovarian endometriotic cysts. Br J Cancer.\n2009;100:1315–9.\n4 2 . G a l g a n oM T ,H a m p t o nG M ,F r i e r s o nHF. Comprehensive analysis of HE4\nexpression in normal and malignant human tissues. Mod Pathol. 2006;19:847–53.\n43. Fortner RT, Vitonis AF, Schock H, et al. Correlates of circulating ovarian cancer\nearly detection markers and their contribution to discrimination of early\ndetection models: results from the EPIC cohort. J Ovarian Res. 2017;10:20.\n44. Urban N, Thorpe JD, Bergan LA, et al. Potential role of HE4 in multimodal\nscreening for epithelial ovarian cancer. J Natl Cancer Inst. 2011;103:1630 –4.\n45. Abdel-Azeez HA, Labib HA, Sharaf SM, Refai AN. HE4 and mesothelin: novel\nbiomarkers of ovarian carcinoma in patients with pelvic masses. Asian Pac J\nCancer Prev. 2010;11:111 –6.\n46. Holcomb K, Vucetic Z, Miller MC, Knapp RC. Human epididymis protein 4 offers\nsuperior specificity in the differentiation of benign and malignant adnexal\nmasses in premenopausal women. Am J Obstet Gynecol. 2011;205:358.e1–6.\n47. Nolen B, Velikokhatnaya L, Marrangoni A, et al. Serum biomarker panels for\nthe discrimination of benign from malignant cases in patients with an\nadnexal mass. Gynecol Oncol. 2010;117:440 –5.\n48. Moore RG, Brown AK, Miller MC, et al. The use of multiple novel tumor\nbiomarkers for the detection of ovarian carcinoma in patients with a pelvic\nmass. Gynecol Oncol. 2008;108:402 –8.\n49. Andersen MR, Goff BA, Lowe KA, et al. Use of a symptom index, CA125, and\nHE4 to predict ovarian cancer. Gynecol Oncol. 2010;116:378 –83.\n50. Anastasi E, Granato T, Falzarano R, et al. The use of HE4, CA125 and CA72-4\nbiomarkers for differential diagnosis between ovarian endometrioma and\nepithelial ovarian cancer. J Ovarian Res. 2013;6:44.\n51. Zhao T, Hu W. CA125 and HE4: measurement tools for ovarian Cancer.\nGynecol Obstet Investig. 2016;81:430 –5.\n52. Goff BA, Agnew K, Neradilek MB, et al. Combining a symptom index, CA125\nand HE4 (triple screen) to detect ovarian cancer in women with a pelvic\nmass. Gynecol Oncol. 2017;147:291 –5.\n53. Jacobs I, Oram D, Fairbanks J, et al. A risk of malignancy index incorporating\nCA 125, ultrasound and menopausal status for the accurate preoperative\ndiagnosis of ovarian cancer. Br J Obstet Gynaecol. 1990;97:922 –9.\n54. Van Holsbeke C, Van Calster B, Bourne T, et al. External validation of\ndiagnostic models to estimate the risk of malignancy in adnexal masses.\nClin Cancer Res. 2012;18:815 –25.\n55. Meys EMJ, Kaijser J, Kruitwagen RFPM, et al. Subjective assessment versus\nultrasound models to diagnose ovarian cancer: a systematic review and\nmeta-analysis. Eur J Cancer. 2016;58:17 –29.\n56. Van Gorp T, Veldman J, Van Calster B, et al. Subjective assessment by\nultrasound is superior to the risk of malignancy index (RMI) or the risk of\novarian malignancy algorithm (ROMA) in discriminating benign from\nmalignant adnexal masses. Eur J Cancer. 2012;48:1649 –56.\n57. Al Musalhi K, Al Kindi M, Al Aisary F, et al. Evaluation of HE4, CA-125, risk of\novarian malignancy algorithm (ROMA) and risk of malignancy index (RMI) in\nthe preoperative assessment of patients with adnexal mass. Oman Med J.\n2016;31:336–44.\n58. Chopra S, Vaishya R, Kaur J. An evaluation of the applicability of the risk of\nmalignancy index for adnexal masses to patients seen at a tertiary Hospital\nin Chandigarh, India. J Obstet Gynaecol India. 2015;65:405 –10.\n59. Campos C, Sarian LO, Jales RM, et al. Performance of the risk of malignancy\nindex for discriminating malignant tumors in women with adnexal masses.\nJ Ultrasound Med. 2016;35:143 –52.\n60. Anton C, Carvalho FM, Oliveira EI, et al. A comparison of CA125, HE4, risk\novarian malignancy algorithm (ROMA), and risk malignancy index (RMI) for\nthe classification of ovarian masses. Clinics. 2012;67:437 –41.\n61. Manegold-Brauer G, Buechel J, Knipprath-Mészaros A, et al. Improved\ndetection rate of ovarian Cancer using a 2-step triage model of the risk of\nmalignancy index and expert sonography in an outpatient screening\nsetting. Int J Gynecol Cancer. 2016;26:1062 –9.\n62. Moore RG, McMeekin DS, Brown AK, et al. A novel multiple marker bioassay\nutilizing HE4 and CA125 for the prediction of ovarian cancer in patients\nwith a pelvic mass. Gynecol Oncol. 2009;112:40 –6.\n63. Chudecka-G łaz AM. ROMA, an algorithm for ovarian cancer. Clin Chim Acta.\n2014;440C:143–51.\n64. Li F, Tie R, Chang K, et al. Does risk for ovarian malignancy algorithm excel\nhuman epididymis protein 4 and CA125 in predicting epithelial ovarian\ncancer: a meta-analysis. BMC Cancer. 2012;12:258.\n65. Kaijser J, Van Belle V, Van Gorp T, et al. Prognostic value of serum HE4 levels\nand risk of ovarian malignancy algorithm scores at the time of ovarian\ncancer diagnosis. Int J Gynecol Cancer. 2014;24:1173 –80.\n66. Wei SU, Li H, Zhang B. The diagnostic value of serum HE4 and CA-125 and\nROMA index in ovarian cancer. Biomed Rep. 2016;5:41 –4.\n67. Sandri MT, Bottari F, Franchi D, et al. Comparison of HE4, CA125 and ROMA\nalgorithm in women with a pelvic mass: correlation with pathological\noutcome. Gynecol Oncol. 2013;128:233 –8.\n68. Plebani M. HE4 study group. HE4 in gynecological cancers: report of a\nEuropean investigators and experts meeting. Clin Chem Lab Med. 2012;50:\n2127–36.\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 8 of 9\n\n69. Karlsen MA, Høgdall EVS, Christensen IJ, et al. A novel diagnostic index\ncombining HE4, CA125 and age may improve triage of women with\nsuspected ovarian cancer - an international multicenter study in women\nwith an ovarian mass. Gynecol Oncol. 2015;138:640 –6.\n70. Chudecka-G łaz A, Cymbaluk-P łoska A, Jastrz ę bska J, Menkiszak J. Can ROMA\nalgorithm stratify ovarian tumor patients better when being based on\nspecific age ranges instead of the premenopausal and postmenopausal\nstatus? Tumour Biol. 2016;37:8879 –87.\nDochez et al. Journal of Ovarian Research           (2019) 12:28 Page 9 of 9","source_license":"CC0","license_restricted":false}