CYFRA 21-1 serum levels in women with adnexal masses and inflammatory diseases.

The aim of the present study was to evaluate the clinical usefulness of the cytokeratin marker CYFRA 21-1 as a screening marker for ovarian cancer, as a predictive marker in patients with adnexal masses and as a prognostic marker in women suffering from ovarian cancer. In order to determine the specificity of the CYFRA 21-1 test, we have investigated CYFRA 21-1 serum levels in several benign conditions. This retrospective study comprises 37 patients suffering from ovarian cancer FIGO stages Ia-III. Sera from patients with benign ovarian cysts, endometriosis, pelvic inflammatory disease, inflammatory bowel disease and liver cirrhosis were evaluated in 90, 10, 38, 10 and 20 cases respectively. With a sensitivity of 41% and a specificity of 95%, CYFRA 21-1 was not suitable as a screening marker for ovarian cancer. Although CYFRA 21-1 was able to discriminate between ovarian cancer and benign adnexal tumours (univariate regression model, P = 0.0001), CYFRA 21-1 did not reveal additional information to CA 125 in a multivariate regression analysis (P = 0.06). CYFRA 21-1 serum levels were elevated in benign conditions such as liver cirrhosis, but not in endometriosis and inflammatory diseases. In ovarian cancer patients, elevated CYFRA 21-1 serum levels before therapy were associated with a poor overall and disease-free survival (log-rank test, P = 0.02 and log-rank test, P = 0.005 respectively). CYFRA 21-1, while obviously not suitable for screening or differential diagnosis of adnexal masses, could be useful as an additional prognostic factor in ovarian cancer patients.

Intermediate filaments consist of fixve different classes: desmin. v-imentin. glial filaments. neurofilaments and cytokeratins (CKs) (Bormer et al. 1994). Cvtokeratins are structural elements of the cytoskeleton of both normal epithelia and their malignant counterparts. Twenty different CKs hax e been identified in human epithelial tissues and have been divided into two subfamilies. basic  and acidic (CK 9-20) CKs. In the surface epithelium of the ovarx. CKs 7. 8. 18 and 19 are found (Moll et al. 1982: Bodenmuller et al. 1994).
CKs of oxarian cancer cells and normal lining cells are identical. but there is a x ast difference in quantity. as intermediate filaments are expressed in relatixely high concentrations in stronglI proliferating, tissues (Moll et al. 1983: Sundstrom et al. 1994 Although the biochemical pathwayvs by which soluble CK fragments are formed in malignant tissues and released into the circulation are not fullN understood. CK serum lex els have been widely reported to be useful indicators of tumour activity in several human malignancies. e.g. colorectal. prostate. breast. cervical. pancreatic and lung, cancer (Gion et al. 1994: Kainz et al. 1994: Plebani et al. 1993).
The serum tumour marker CYFRA 21-1 detects a fragment of CK 19 (Pujol et al. 1995). Basically. CK 19 is expressed in all epithelia and epithelia-derixved tissues. but it has also been detected in other cell types. e.g. peripheral blood mononuclear cells Received 5 December 1997Revised 16 March 1998 Accepted 24 March 1998 Correspondence to: C Tempfer. Department of Gynaecology and Obstetrics. Vienna University Medical School, A-1090 Vienna. Wahringer GCirtel 18-20. Austna (Novaes et al. 1997). CK 19 has been shoxxn to be a marker of cytodifferentiation in the human fetal endocrine pancreas (Bouwens et al. 1997). Elevated serum levels of CK 19 have been descnrbed in v-arious human malignancies. e.g. pancreatic. bladder and cervical cancer (Kainz et al. 1995: Senga et al. 1996: Grem et al. 1997. CYFRA 2 1-1 has been shoxwn to be a clinicallv valuable prognostic and monitoringc marker in oesophageal cancer. head and neck cancer and in non-small-cell luncg cancer (NSCLC) (Plebamu et al. 1995: Goumas et al. 1997: Yamamoto et al. 1997 (Mobus et al. 1992: Yanagibashi et al. 1997. To the authors' knoxwledge. no data on serum levels of CYFRA 2 1-1 in ox arian cancer have been reported so far. The aim of the present studv w-as to exvaluate xhether CYFRA 21-1. alone or in combination wxith CA 125. has a potential as a screening marker in epithelial oxarian cancer. Furthermore. we have compared preoperative serum lex els of CYFRA 2 1-1 and CA 125 in patients with benign ovarian cysts and epithelial oxvarian cancer. wxith recard to their value in the differential diarnosis of adnexal masses. Additionallv. w e hax-e inxestigated the prornostic potential of CYFRA 21-1 serum lexels evaluated before therapy. Serum lexels of cytokeratins are known to be elexated in several benign conditions. e.g. inflammatorv diseases (Ouhavoun et al. 1990: Nakamura et al. 1997) and lixer disease (Sabbatini et al. 1988). To identify possible benign conditions being associated with elex ated CYFRA 21-1 lexvels. x e have evaluated serum samples of patients with endometriosis. pelx ic inflammatordisease. inflammator bowel disease. lixer cirrhosis and hepatitis.

MATERIALS AND METHODS
This retrospective study includes 175 preoperative serological examinations of patients with clinically doubtful findings in the small pelvis. Thirty-seven of them were suffering from ovarian cancer FIGO stages Ia (n = 2). Ic (n = 5). II (n = 8) and III (n = 22).
Median age at the time of diagnosis was 57.9 (range 29-87) years.
Histologically. 12 tumours were graded as serous adenocarcinoma. nine as mucinous adenocarcinoma. five as undifferentiated carcinoma. three as endometrioid carcinoma. one as clear cell carcinoma. and seven as other kinds of oxarian cancer. Benign cysts. endometriosis and pelvic inflammatorv disease were found in 90. 10 and 38 patients respectivelv.
AdditionalIv. we haxe investigated the sera of ten and 20 patients suffering from inflammatorv bowel disease (Crohn's disease and ulceratixe colitis) and liver cirrhosis respectively. Serum lexvels of CYFRA 21-1 were additionally evaluated in a panel of 40 female blood donors. All of these women were apparently healthy and had no history of malignancy. Median age was 31.3 (range 19-53) years.

Clinical management
All patients sufferinc from ovarian cancer underent total abdominal hysterectomy. bilateral salpingo-oophorectomy. pelvic and para-aortic lymphadenectomy and omentectomy. Patients with stages lc-[II and patients with clear cell carcinoma underwent a platinum-containing chemotherapy regimen. All patients were followed up in 3-month interxals. including vaginorectal palpation. abdominal ultrasound examination and serum tumour marker evaluation. The median duration of follow-up was 21.5 (range 0.5-67) months. Sixteen patients developed progressive/recurrent disease after primary therapy. with a median time to progression of 5 (range 0-14.5) months. Sixteen patients died of the disease.
Serum assay Blood samples were collected by peripheral vein puncture. allowed to clot. centrifuged and stored in four aliquots at -80'C.
Serum concentrations of CYFRA 21-1 were measured using the CYFRA 21-1 enzvme-linked immunosorbent assay (ELISA) (Boehringer Mannheim. Mannheim. Germany). a two-site enzyme immunoassay for the detection of cvtokeratin 19. The assay was carried out according to the manufacturer's instructions. using the automated ELISA processor ES 300 (Boehringer Mannheim). The intra-assay coefficient of correlation was 6.5%7 at a concentration of 3 igo 1-1. Serum CA 125 was measured using, an immunoradiometric assay (Abbott CA 125 RIA Diagnostic Kit. Abott Laboratories. NC. USA). Variation coefficient at 46 U ml for ten assays was 5.2%c. All tests were run in duplicate.

Statistical analysis
Because of their skewed distribution. median xvalues (range) are gixen to describe serum CYFRA 21-1 and CA 125 levels. Logarithmic-transformed values were used for further analysis. Logistic regression models (Hosmer et al. 1989) were used to analyse the influence of serum CYFRA 21-1 and CA 125 levels on the probability of malignancy. Using a logistic regression model. sensitivity and specificity were calculated for each possible threshold value of estimated probabilitv for malignancy. Based on these values. receiver operator characteristics (ROC) curxes (Campbell et al. 1996) were constructed ( Figure IA). A logistic regression model was used to compare healthy xA omen and ovarian cancer patients with respect to their CYFRA 21-1 xalues.
Univariate and multivariate logistic regression analyses were used to compare patients with benign ovarian cysts and ovarian cancer patients with respect to their CYFRA 21-1 and CA 125 values. The ROC curves (Figure 1 B) show the influence of each of the two variables on the probability for malignancy. as well as the diagnostic power of their simultaneous consideration.
Comparison of serum levels between unpaired groups were made using the Mann-Whitney U-test. Sun ival probabilities were calculated by the product limit method of Kaplan and Meier.
Differences betseen groups in survixal curves were assessed
Preoperative CYFRA 21-1 serum levels as prognostic factors in ovarian cancer As CYFRA 21-1 lacks a clearix defined cut-off value. a cut-off x-alue of 9.4 jgc 1-1 was selected according to the 0.75 quantile (upper quartile) of serum concentrations measured in the panel of ox anan cancer patients (Obermair et al. 1997). Using the product limit method of Kaplan and Meier. w-e calculated the probability of pretreatment CYFRA 21-1 serum lexels to predict the overall survival. Elevated CYFRA 21-1 serum levels before therapyxwere associated with a poor overall and disease-free surnix-al (log-rank test. P = 0.02. Fioure 2: and log-rank test. P = 0.005. Figure 3. respectively). Patients with and xxithout elevated CYFRA 21-1 serum levels did not show statistically significant differences regardincg treatment modality. performance status. mean age of the patients and distribution of tumour stage.
CYFRA 21-1 and CA 125 serum levels in benign conditions

DISCUSSION
Although cvtokeratin tumour markers have been investigated in a wide ariety of human malignancies. fexx data on cytokeratins in ovanan cancer exist.
In the present study. serum CYFRA 21-1 showxed a sensitivitx of 41 %7 and a specificity of 95% in ovarian cancer patients. The ROC 1.0o- FKgure 3 Kaplan-Meer analysis regarding disease-free survival of ovanan cancer patents with CYFRA 21-1 serum levels above Mte cut-off level (9.4 jig h1; n = 9) and CYFRA 21-1 serum levels below the cut-off level (9.4 jg 1-1: n = 28) curxe shows that CYFRA ' 1-1 is not suitable as a screening marker for ovarian cancer. Gix en the low prex alence of the disease (50/100 000). the CYFRA 21-1 test would yield only 1 in 245 wxomen with a positive test actuallv havIing the disease. With respect to differential diagnosis of adnexal masses. elevated CYFRA 21-1 serum levels sho%ved a positive correlation with the risk of presenting with malionant disease. The higher the CYFRA 21-1 lexels. the higher was the risk of carrying a malinant ox arian cyst. This points to the fact that cytokeratin release is a genuine part of ovarian cancer dex elopment. Howex er. our study also shows that CYFRA 21-1 does not reveal additional diagnostic information in the presence of the established tumour marker CA 125. This documents that CYFRA 21-1 is not clinically useful as an additional discriminator between the absence or presence of malignancy in patients intended to undergo surgery for suspicious ox arian c sts.
It has to be stressed that all tumour marker results haxe to be interpreted wxith caution. Even CA 125. which is generallv established as a specific tumour marker for oxarian cancer. has been reported to be elexated in xvarious benign conditions as well as in other malignancies. e.g. endometriosis and endometrial cancer (Kramer et al. 1993: Rose et al. 1994. This is also true for cxtokeratins. which haxe been reported to be elevated in a wide xariety of human malignancies (Gion et al. 1994: Sliutz et al. 1995. This fact is based on the abundant expression of cytokeratins. wAhich are found in all epithelia and epithelia-derived tissues. Thus. cvtokeratins or cvtokeratin fragments are expressed in large quantities in all proliferating epithelial tissues. Diseases of the liver hax-e been reported to be associated with elexated serum lexels of cytokeratins (Sabbatini et al. 1988). This is consistent w ith our findings of elex ated CYFRA 21-1 serum levels in patients wxith lix-er cirrhosis. When interpreting elex ated CYFRA 21-1 serum lexels. it has to be taken into account that CYFRA 21-1 is not a specific ovarian cancer tumour marker and that the coincidence of other bemnin diseases or malignancies may lead to elexated serum lexels and consequentlI to false-positix e test results. CK 19 has been described as an indicator of epithelial injuries in inflammatory diseases. e.g. parodontosis and chronic bronchitis (Ouhayoun et al. 1990: Nakamura et al. 1997). In the present study. CYFRA 21-1 was not increased in inflammatory bowel disease or pelxvic inflammatory disease. This indicates that inflammation per se is not sufficient to cause elevated CK levels.
In the present study. we found no correlation between CYFRA 21-1 serum lexvels and tumour staae. This finding supports the assumption that serum lexels of cytokeratins are not reflective of the tumour bulk but rather indicative of strongly proliferating, tumours (Bormer et al. 1994).
Elevated serum levels of cvtokeratins have been shown to provide prognostic information in several malignancies (Kainz et al. 1994: Carpelan-Holmstrom et al. 1996. In our study. we found that increased CYFRA 21-1 serum levels before therapy are predictive of the patients' outcome in oxarian cancer. Patients x ith elexated CYFRA 21-1 serum lexels had a shortened oxerall and disease-free surxvival. Additional studies with an increased number of patients are justified to clarifx further the prognostic value of CYFRA 21-1 in ovarian cancer patients. Healthy controls 40 1.9 (0.1-11) -In conclusion. our data indicate that CYFRA 21-1 is suitable neither as a screening marker nor as an additional tool for differential diagnosis of adnexal masses. Regarding benign diseases. patients with liver cirrhosis show extremely high cytokeratin levels. whereas inflammatory conditions are not invariably associated with elevated serum levels of CYFRA 21-1. In ovarian cancer patients. serum levels of CYFRA 21-1 are not associated with clinicopathological parameters. The most promising result of this study is the prognostic value of preoperative CYFRA 21-1 serum levels regarding overall and disease-free survival. Considering these results. CYFRA 21-1. while obviously not suitable for screening or differential diagnosis of adnexal masses. could be useful as an additional prognostic factor in ovarian cancer patients.