Printer Friendly

Combined inhibin and CA125 assays in the detection of ovarian cancer.

Inhibin is a dimeric glycoprotein produced by the gonads that is involved in the regulation of pituitary follicle-stimulating hormone secretion [for reviews, see Refs. (1-3)]. Inhibin consists of two chains, the [alpha] subunit (made up of three regions, Pro-, [alpha]N, and [alpha]C) and either the [beta]A subunit (to form inhibin A) or [beta]B subunit (inhibin B) of different molecular weights. The nonbiologically active free a subunit and fragments have also been identified in biological fluids, including serum. Earlier studies (4-8) showed that the measurement of serum inhibin by RIA, which detects all [alpha]C-containing inhibin forms together with free [alpha] subunit, was of value in the diagnosis and monitoring of ovarian granulosa cell and mucinous tumors in women after menopause.

An [alpha]C immunofluorometric assay (IFMA) [6] for inhibin was developed recently (9,10) to match the assay specificity of the RIA, but with the advantages of increased sensitivity and speed. This assay detects all known inhibin [alpha] subunit-containing proteins. This assay has been used to characterize inhibins in fractionated human follicular fluid and serum from healthy men and women (9,10), and showed a specificity similar to the RIA in detecting the various inhibin forms.

The aims of this study were to validate the [alpha]C IFMA in application to serum and then to compare its diagnostic accuracy with the RIA in the detection of ovarian mucinous and granulosa cell cancers. In addition, the value of combining inhibin measurements with that of CA125, which readily detects serous tumors (11-13), in the detection of all ovarian cancers was explored.

Materials and Methods

SUBJECTS

Serum samples were obtained preoperatively from 177 women admitted to major gynecologic teaching hospitals in Melbourne with symptoms consistent with ovarian cancer and who subsequently were shown to have ovarian (n = 154) or nonovarian (n = 23) cancers. Classification of the tumors was similar to that used previously (5).

The various cancer types investigated are presented in Table 1. Nine of 11 women with granulosa cell tumors were [greater than or equal to]50 years of age, but two were 30-40 years of age and had undergone bilateral salpingo-oophorectomy before the study. Serum samples were obtained from 165 healthy postmenopausal women attending a mammography breast screening clinic. The women studied were asymptomatic and considered postmenopausal based on age ([greater than or equal to]55 years) or increased serum follicle-stimulating hormone (>20 IU/L). The research and ethics committees of all participating hospitals approved the study protocol.

INHIBIN RIA

Inhibin was measured with a heterologous RIA (14), using iodinated 31-kDa inhibin as tracer and a rabbit antiserum (no. 1989) raised against bovine 31-kDa inhibin. This assay showed cross-reaction with inhibin, Pro-[alpha]C, and Pro-[alpha]N-[alpha]C, but not with [alpha]N fragment, activin, or follistatin (10). The RIA calibrator used was a human serum pool (designated MMC-1) of arbitrary units. The RIA sensitivity between assays was 35-110 units/L. This assay has been used previously for measuring inhibin in serum from postmenopausal women with ovarian cancer (4,5). The serum inhibin value (122 units/L) used to discriminate between control and cancer groups in the RIA was determined previously (5). The majority (>90%) of control samples gave nondetectable values in this RIA. The MMC-1 inhibin calibrator when assayed against the First International Standard for inhibin A, (human recombinant, 30-34 kDa, 91/624; National Institute of Biological Standards and Control, Potters Bar, UK) in the [alpha]C IFMA was 1.80 ng of 91/624 per unit of MMC-1 with 95% confidence limits of 1.54 and 2.09 ng/unit.

[alpha]C SUBUNIT IFMA

A two-site IFMA was developed with the aim of detecting all [alpha]C subunit-containing inhibin forms in serum. The capture antibody used was a caprylic acid IgG cut of a sheep polyclonal antibody (no. 128) raised initially against human inhibin [alpha]C subunit fusion protein (10) and subsequently boosted with recombinant inhibin A. The antibody used as label was a sheep polyclonal antibody (no. 41) raised against human inhibin [alpha]C subunit fusion protein (10). The IgG fraction was immunopurified by fractionation on a column of immobilized bovine inhibin [alpha]C-subunit fusion protein and eluted with 4 mol/L guanidine hydrochloride. The guanidine hydrochloride was removed by gel filtration on Sephadex G25 columns (PD-10 column; Pharmacia) and gel filtration HPLC. The IgG fraction was concentrated by a centrifugation procedure (Centricon; Amicon) to 1 g/L and biotinylated (4 h at room temperature) using the biotin-isothiocyanate procedure (Wallac; Turku) in 50 mmol/L [Na.sub.2]C[O.sub.3], buffer, pH 9.8, tested over a 60- to 150-fold molar excess of biotin reagent. The sheep antisera and fusion protein were gifts from Biotech Australia P/L, Sydney, Australia.

The assay was carried out in 96-well microtiter plates (Maxisorb; Nunc). Wells were coated with 2 [micro]g/well capture antibody in glycine buffer, pH 4.4, and blocked with 50 mmol/L Tris-HCI, pH 7.5, containing 10 mg/L bovine serum albumin (BSA) for 1 day at room temperature. The assay buffer used was 50 mmol/L Tris-HCI, pH 7.4, containing 0.154 mol/L NaCl, 1 g/L [NaN.sub.3], and 5 g/L BSA (TSA-BSA) containing 1 g/L ovine IgG. To reduce the matrix effects of serum in the assay, we diluted the calibrator and serum samples in serum from which the inhibin had been removed by adsorption to an immobilized no. 41 antiserum IgG fraction. The serum was initially recycled five times through the immunoaffinity support, although two cycles were sufficient to remove all inhibin immunoactivity as measured in the [alpha]C IFMA. The effectiveness of the depletion process was assessed from the low stable counts obtained in the [alpha]C IFMA with repeated cycles through the immunoaffinity support and the similar responses observed for the depleted serum and assay buffer alone in the [alpha]C IFMA. Serum samples (100 [micro]L) or calibrators (100 [micro]L) and TSA-BSA buffer (100 [micro]L) were incubated in the antibody-coated microtiter plate for 2 h at room temperature with shaking. The plate was then washed, and biotinylated antibody (150 ng/ well) was added and incubated an additional 2 h at room temperature. After the plate was washed, Eu-streptavidin (50 ng/well; Wallac) was added; the plate was incubated for 30 min and then washed six times, and enhancement solution (Delfia; Wallac) was added. The plates were read on a Wallac 1234 Fluorometer. The dose-response curves were calculated using the Multicalc program (Wallac). Where appropriate, potencies and evidence of parallelism between serial dilutions of calibrator and serum samples were determined using parallel line statistics as outlined in Finney (15)

The inhibin A (91/624) preparation was used as calibrator and expressed in terms of its nominal vial content (5 [micro]g). The specificity of the [alpha]C IFMA had been assessed previously (10). The [alpha]C IFMA detected all the known molecular weight forms of inhibin. The detection limit of the assay represented the mean inhibin value detected 2 SD above the zero value of the [alpha]C IFMA from five experiments.

CA125

CA125 was measured by an immunoenzymometric assay (AIA-PACK CA125; TOSOH) using a CA125 calibrator provided by the manufacturer. The manufacturer stated that the mean serum CA125 concentration in 172 apparently healthy individuals was 12.3 kilounits/L, with an upper limit of 33.4 kilounits/L.

CA125 and inhibin by RIA were determined by the Monash Medical Centre Chemical Pathology Department. [alpha]C IFMA measurements were undertaken at Prince Henry's Institute of Medical Research, Melbourne.

GENERAL STATISTICAL ANALYSES

For statistical purposes, the patient sera were divided into five main groups: controls, serous, mucinous, granulosa cell, and miscellaneous. The cancers of nonovarian origin were considered separately. In addition, the four ovarian cancer groups were combined to form an all-ovarian cancer group. The data were normalized after log transformation as confirmed using normal probability plots, half-normal plots, and histograms. The data were log-transformed before analysis. The 5th and 95th percentiles for each group were determined as the antilogs of the 5th and 95th percentiles of the normally distributed log-transformed data.

ROC CURVES

The ROC curve (16,17) for a diagnostic test plots the sensitivity against the false-positive rate. In this context, the sensitivity is the proportion of cancer patients who are identified correctly by the diagnostic test, so-called true positives; the specificity is the proportion of control patients who are identified correctly, so-called true negatives. One minus the specificity equals the false-positive rate. The ROC curve can be used to select a cutoff for the diagnostic test that maximizes the sensitivity and minimizes the false-positive rate. The areas under the ROC curves can be used to compare the performance of the two assays.

CANONICAL VARIATE ANALYSIS

The data from different assays can be combined to improve the sensitivity and specificity relative to those of the individual assays. One approach to finding an optimal combination of several assays is canonical variate analysis (CVA) (18). The optimal solution chosen by CVA is the linear combination of assays that maximizes the F-statistic of between-group variance divided by within-group variance, where group identifies the cancer and control patients.

MULTIROC ANALYSIS

The multiROC, like CVA, is a method (19) for combining the results of multiple (two or more) assays through the derivation of a composite test rule. For example, if two assays are available, then a patient is classified as a cancer patient if the value for the first assay is greater than the cutoff for that assay or the corresponding value for the second assay is greater than its corresponding cutoff value. The multiROC curve is drawn by constructing the ROC curve for the first assay in the usual way, selecting the "best" cutoff for the first assay, and then adding another curve for the second assay, which completes the multiROC for the composite rule.

COMPARISON OF CVA AND MULTIROC

The multiROC is an ingenious method, but has some difficulties. First, although the cutoffs for the individual assays can be "optimally" derived from ROC curves, there is no guarantee that those same cutoffs will be the optimal cutoffs when a composite decision rule is formed. Second, although graphical techniques have been proposed to aid in the formulation of a decision rule for the multiROC, the search for the optimal rule and optimal set of cutoffs is subjective and can become difficult as the number of assays increases.

CVA is not a panacea for these difficulties, but it does have some advantages and, therefore, is worthy of consideration: (a) the method is mathematically optimal in the sense of maximizing the F-statistic; (b) the method is objective; and (c) the method can be extended easily to multiple assays.

Results

VALIDATION OF THE INHIBIN [alpha]C IFMA IN APPLICATION TO SERUM

A two-site IFMA was developed as a more sensitive and rapid method to replace the RIA (14) with the objective of detecting all known [alpha]C subunit-containing proteins in serum. The inhibin standard (WHO inhibin A reference preparation, 91/624) gave a curvilinear response with a detection limit of 2 pg/well or 20 ng/L sample (Fig. 1) and an upper response >5 ng inhibin/well (data not shown). Inhibin-free serum was included to reduce serum matrix effects in the assay. Ovine IgG at 1 g/L was added to eliminate the effects of antibody cross-linking agents in human serum. In a preliminary series, ~10% of sera showed responses that were eliminated after the inclusion of ovine IgG.

[FIGURE 1 OMITTED]

Serial dilutions of a human serum pool gave a parallel response to the dose-response curve of the inhibin standard (Fig. 1). The detection limit of the assay based on a response 2 SD above the controls was 2 pg/well or 20 ng/L serum. The mean precision profile of five calibration curves is presented in Fig. 1. The within-assay imprecision (CV) as calculated from the reproducibility of a serum pool (69 [+ or -] 5.5 pg/well) used as quality control in the present studies was 4.2% (n = 4), and the between-assay CV was 8% (n = 4).

The [alpha]C IFMA and CA125 were applied to serum samples from 165 healthy postmenopausal women, 154 postmenopausal women with a variety of ovarian cancers, and 23 women with nonovarian cancers (Table 1). The RIA was applied to the cancer samples only, because previous studies (5) had shown that >90% of serum samples from healthy postmenopausal women were non-detectable in this assay.

The data from the three assays, expressed as geometric means with 5th and 95th percentiles, were grouped by cancer class into seven main groups: controls, serous cancers, mucinous cancers, and granulosa cell cancers, a miscellaneous series of ovarian cancers, an all-ovarian cancer group, and nonovarian cancers (Table 1). The miscellaneous cancer group has been subgrouped into individual cancers and presented in Table 1B.

COMPARISON OF THE VARIOUS CANCER GROUPS BETWEEN ASSAYS

The specificity and sensitivity of the [alpha]C IFMA, RIA, and CA125 were assessed initially from the number of cancer and control values in each assay that were detected above the 95th percentile of the geometric mean of the control group, or in the case of the RIA, a previously determined cutoff point (122 units/L, Table 1). On the basis of this approach, the [alpha]C IFMA showed similar or better discrimination characteristics compared with the RIA in the detection of mucinous (90% vs 71%, respectively) and granulosa cell tumors (100% vs 100%). In contrast, the serous and miscellaneous tumors were better discriminated by the CA125 method.

The [alpha]C IFMA and CA125 were also assessed by ROC curve analysis. The ROC curves are presented in Fig. 2 for the major cancer groups. The areas under the ROC curves for each analysis and statistical comparisons of areas under the curves are presented in Table 2. These data in support of the results above show that at 95% specificity, [alpha]C IFMA detected 95% of mucinous and 100% granulosa cell tumors, whereas CA125 detected 92% of serous and 79% of miscellaneous cancers.

COMBINATION OF ASSAYS FOR THE VARIOUS CANCERS

CVA/ROC curve and multiROC analyses were undertaken to establish whether the combination of IFMA and CA125 assays leads to an improved accuracy in the detection of all ovarian cancers. Application of both methods in assessing the combination of the CA125 and IFMA led to the detection of a high proportion of all ovarian cancers. This is exemplified in Table 2, where the areas under the ROC curves for the CA125 + [alpha]C IFMA combination for the all-ovarian cancer group was significantly greater (P <0.05) by either the CVA (Fig. 3, top) or the multiROC (Fig. 3, bottom) method in comparison with either assay alone. The percentage of all ovarian cancers detected at 95% specificity by the combined assays was 89-90% vs 80% for CA125; at 90% specificity, the percentage of ovarian cancers detected was 91-93% vs 84% (Table 3).

[FIGURE 2 OMITTED]

The relationship between CA125 and the [alpha]C IFMA and their combination in differentiating between controls and ovarian cancers is presented as a scatter plot in Fig. 4. The control group is confined to the lower left-hand quadrant, separate from the ovarian cancer groups in the other quadrants. The vertical line corresponds to the 95% specificity value (28.5 units/L CA125; see Table 2 and Fig. 3) obtained using a ROC analysis and was chosen as the cutoff for CA125 for the multiROC decision rule. On the basis of the multiROC graph in Fig. 3, we decided to increase the [alpha]C IFMA value from 129 ng/L (Table 2) to 218 ng/L, shown as the horizontal line in Fig. 4. Thus, the multiROC decision rule was to classify a patient as a case if the result obtained was >28.5 units/L in the CA125 or was >218 ng/L in the [alpha]C IFMA. The diagonal line was derived by CVA and represents the sensitivity values at 95% specificity. As seen in Table 2, the sensitivity values at 95% specificity of the combined CA125 + [alpha]C IFMA analyses are similar when either the CVA or multiROC curve analysis is used.

Regression analysis between inhibin assays was assessed for the various cancer groups. As seen in Table 4, regression analyses between [alpha]C IFMA and RIA showed high correlations (r = 0.63-0.83) for mucinous and granulosa cell cancer groups and all-ovarian cancer group with low correlations (r = 0.31-0.5) for the other cancer groups.

Discussion

It had been shown previously (4,5) that serum inhibin as measured by RIA is a good marker for mucinous and granulosa cell tumors--cancers that are not readily detected by CA125. However, the limited sensitivity of the RIA as reflected by its failure to detect inhibin in the majority of healthy postmenopausal sera suggested that more sensitive inhibin assays might provide a better discriminatory response. Nonetheless, because it is not known which inhibin forms are present in ovarian cancer serum, the appropriate or required specificity of the proposed new inhibin assay was unclear. The inhibin RIA detects all molecular weight forms of dimeric inhibin as well as the free a subunit (9,10). Previous studies (7, 20, 21) showed that inhibin A was detected in 20% of mucinous tumors and 77% of granulosa cell tumors, whereas inhibin B was detected in 60% of mucinous tumors (21) and 100% of granulosa cell tumors. Similarly, Pro-[alpha]C, which is a monomeric inhibin [alpha] subunit form, was detected by Pro-[alpha]C ELISA (22) in 90% of granulosa cell cancers and 55% of mucinous cancers (21). These findings indicate that a replacement assay for the RIA should show a broad range of specificity capable of detecting free [alpha] subunit forms as well as dimeric inhibin forms. The [alpha]C IFMA satisfies these requirements by detecting all known inhibin forms containing the [alpha]C region of the a subunit as well as the free [alpha] subunit with an assay specificity which closely parallels the RIA (9,10), thus providing a similar broad specificity range. The [alpha]C IFMA is likely to detect the presence of degraded forms (9,10) because it detects inhibin denatured by reducing agents, whereas the RIA does not (14).

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

The [alpha]C IFMA shows a similar or improved detection of mucinous and granulosa cell tumors compared with the RIA used previously. This improvement may be attributed to the increased sensitivity of the IFMA, in which inhibin is detected in 84% of healthy samples compared with the limited proportion detected by the RIA.

The observation that CA125 is effective in detecting serous and miscellaneous cancer groups, whereas inhibin is effective in detecting the mucinous and granulosa cell cancer groups suggests that these two analytes may provide an appropriate combined test for detecting the majority of ovarian cancers. A combination of the CA125 and the [alpha]C IFMA by canonical variate/multiROC analysis was thus undertaken. The data from the individual and combined assays were analyzed either on the basis of the proportion of cancers detected above the 95th percentile of the controls (which is used to analyze the RIA data) or by ROC analysis. ROC analysis permits the assessment of sensitivity at any specificity value; however, the data have been considered as an example at 95% specificity. The analysis showed that the combination of CA125 and [alpha]C IFMA detected a significantly higher proportion of all ovarian cancers compared with either assay alone. At 95% specificity, that is, at a discrimination point where 5% of healthy controls are detected, the sensitivity of cancer patients in the all-cancer group was 89-90%, compared with 80% with CA125 or 50% with aC IFMA alone. This observation suggests that in the clinical assessment of ovarian cancers, the combination of inhibin and CA125 increases the probability that the majority of ovarian cancers will be detected.

It was anticipated that the combination of the two assays would lead to a synergistic response that would allow the detection of cancers with subthreshold values in the two assays. A comparison of the CVA method with the multiROC method yields a similar discrimination index, indicating that this is not the case. One explanation is that the individual assays are particularly effective in detecting their respective cancers with little overlap between cancer groups. Because no synergism was observed, a positive response in one or both assays will provide evidence of cancer without the need for an appropriate algorithm.

Other markers in combination with CA125 have been reported to improve their diagnostic accuracy in the pretreatment evaluation of ovarian cancers. Stenman et al. (23) reviewed the sensitivity/ specificity characteristics of several markers and recommended CA125 and either tumor-associated trypsin inhibitor or the gastrointestinal mucin markers CA 19-9 or CA 72.4 for mucinous tumors. The combination of CA125 and tumor-associated trypsin inhibitor for mucinous tumors gave 100% sensitivity at 89% specificity (24), which is similar to that found with CA125 + [alpha]C IFMA (100% sensitivity and 90% specificity) in this study. However, tumor-associated trypsin inhibitor is less effective in detecting granulosa cell tumors. CA125 + CA 72.4 gave sensitivity values of 85% and 86% and specificities of 81% and 86% for two studies examining all ovarian cancers (25, 26). These values are lower than the corresponding values (95-100% sensitivity and 95% specificity) obtained in the present study. These findings suggest that the combination of CA125 and [alpha]C IFMA provides a better biochemical diagnosis of all ovarian cancers than other available tests or combination of tests. It should be noted that the CA125 + aC IFMA combination applies to postmenopausal women, whereas the other combinations are also applicable to premenopausal women.

The tissue specificity of the [alpha]C IFMA was assessed by examination of a series of nonovarian cancers that gave limited detection similar to that obtained with the RIA (5).

A previous study (5) examined the RIA inhibin concentrations in 33 benign ovarian tumors, 11 of which (3 of 4 fibromas, 4 of 4 thecomas, and 2 of 4 mucinous cystadenomas) had increased inhibins. A corresponding group was not assessed in the present study, but it is likely that a similar distribution would be found with other inhibin a-subunit assays. The specificity of the assay in application to cancers of nonovarian origin showed a low discrimination index (17%, Table 1A). Nonetheless, studies of a wider range and number of nonovarian cancers and other disease processes will be needed to establish the specificity of the assay in its clinical application.

The use of inhibin as a marker of ovarian disease is limited to women after menopause, when their serum inhibin concentrations are low. Serum inhibin concentrations produced by the ovary during the reproductive years show substantial fluctuations that would seriously compromise a preoperative assessment of any ovarian cancer. However, inhibin measurements may be useful post surgery in premenopausal ovarian cancer patients during the recovery/ recurrence phase because these women undergo oophorectomy as the first stage of their treatment. Inhibin should be considered the marker of first choice for monitoring treatment of women with mucinous ovarian cancers. Studies are directed toward establishing whether there are cancer specific inhibin forms, which may be suitable during the reproductive years.

This study was funded by program grants (943208 and 983212) of the National Health and Medical Research Council of Australia. We thank Biotech Australia (Sydney, Australia) for providing the inhibin A and the sheep antisera, and Thoa Nguyen for assistance in the statistical analysis.

Received September 28, 1998; accepted March 3, 1999.

References

(1.) Vale WW, Hseuh A, Rivier C, Yu J. The inhibin/activin family of hormones and growth factors. In: Sporn MA, Roberts AB, eds. Peptide growth factors and their receptors. Handbook of experimental physiology. Vol. 95. Berlin: Springer-Verlag, 1990:211-48.

(2.) Burger HG. Inhibin. Reprod Med Rev 1992;1:1-20.

(3.) Baird DT, Smith KB. Inhibin and related peptides in the regulation of reproduction. Oxf Rev Reprod Biol 1993;15:191-232.

(4.) Lappohn RE, Burger HG, Bouma J, Bangah M, Krans M, de Bruijn HWA. Inhibin as a marker for granulosa cell tumours. N Engl J Med 1989;321:790-3.

(5.) Healy DL, Burger HG, Mamers P, Jobling T, Bangah M, Quinn M, et al. Elevated serum inhibin concentrations in postmenopausal women with ovarian tumors. N Engl J Med 1993;329:1539-42.

(6.) Cooke I, O'Brien MO, Charnock FM, Groome N, Ganesan TS. Inhibin as a marker for ovarian cancer. J Cancer 1995;71:1046-50.

(7.) Burger HG, Robertson DM, Cahir N, Mamers P, Healy DL, Jobling T, Groome N. Characterization of inhibin immunoreactivity in post-menopausal women with ovarian tumors. Clin Endocrinol 1996;44:413-8.

(8.) Lambert-Messerlian GM, Steinhoff M, Zheng W, Canick JA, Gajewski WH, Seifer DB, Schneyer A. Multiple immunoreactive inhibin proteins in serum from postmenopausal women with epithelial ovarian cancer. Gynecol Oncol 1997;65:512-6.

(9.) Robertson D, Burger HG, Sullivan J, Cahir N, Groome N, Poncelet E, et al. Biological and immunological characterisation of inhibin forms in human plasma. J Clin Endocrinol Metab 1996;81:669-76.

(10.) Robertson DM, Cahir N, Findlay JK, Burger HG, Groome N. The biological and immunological characterisation of inhibin A and B forms in human follicular fluid and plasma. J Clin Endocrinol Metab 1997;82:889-96.

(11.) Bast RC Jr, Klug TL, St John E, Jenison E, Niloff JM, Lazarus H, et al. A radioimmunoassay using a monoclonal antibody to monitor the course of epithelial ovarian cancer. N Engl J Med 1983;309: 883-7.

(12.) Bast RC Jr, Knauf S, Epenetos A, Dhokia B, Daly L, Tanner M, et al. Coordinate elevation of serum markers in ovarian cancer but not benign disease. Cancer 1991;68:1758-63.

(13.) van der Burg MEL, Lammes FB, Verweij J. CA125 in ovarian cancer. Neth J Med 1992;40:36-51.

(14.) Robertson DM, Tsonis CG, McLachlan RI, Handelsman DJ, Leask R, Baird DT, et al. Comparison of inhibin immunological and in vitro biological activities in human serum. J Clin Endocrinol Metab 1988;67:438-48.

(15.) Finney DJ. Statistical method in biological assay. London: Griffin, 1971:668 pp.

(16.) Hanley JA, McNeil BJ. The meaning and use of the area under a receiving operating characteristic (ROC) curve. Radiology 1982; 143:29-36.

(17.) Hanley JA, McNeil BJ. A method of comparing the areas under receiving operating characteristic curves derived from the same cases. Radiology 1983;148:839-43.

(18.) Krzanowski WJ. Principles of multivariate analysis. Oxford: Clarendon Press, 1988:289-306.

(19.) Shultz EK. Multivariate receiver-operating characteristic curve analysis: prostate cancer screening as an example. Clin Chem 1995;41:1248-55.

(20.) Petraglia F, Luisi S, Pautier P, Sabourin J-C, Rey R, Lhomme C, Bidart J-M. Inhibin B is the major form of inhibin-activin family secreted by granulosa cell tumours. J Clin Endocrinol Metab 1998;83:1029-32.

(21.) Robertson DM, Cahir N, Burger HG, Mamers P, Groome N. Inhibin forms in serum from postmenopausal women with ovarian cancers. Clin Endocrinol (in press).

(22.) Groome NP, Illingworth PJ, O'Brien M, Priddle J, Weaver K, McNeilly AS. Quantitation of inhibin Pro-aC-containing forms in human serum by a new ultrasensitive two-site enzyme-linked immunosorbent assay. J Clin Endocrinol Metab 1996;80:2926-32.

(23.) Stenman U-H, Alfthan H, Vartiainen J, Lehtovirta P. Markers supplementing CA125 in ovarian cancer. Ann Med 1995;27:115-20.

(24.) Mogensen O, Megensen B, Jakobsen A. Tumour-associated trypsin inhibitor (TATI) and cancer antigen 125 (CA 125) in mucinous ovarian tumours. Br J Cancer 1990;61:327-9.

(25.) Malvano R, Ferdeghini M, Chicchio A, et al. Tumour markers in epithelial ovarian cancer: approaches to the interpretation of laboratory data. In: Ballesta A, Torre C, Bombardieri E, Gion M, Molina R, eds. Updating on tumour markers in tissues and in biological fluids. Torino: Edizione Minerva Meddica, 1993: 725-35.

(26.) Gadducci A, Ferdeghini M, Prontera C, Moretti L, Mariani G, Bianchi R, Fioretti P. The concomitant determination of different tumour markers in patients with epithelial cancer and benign ovarian masses: relevance for differential diagnosis. Gynecol Oncol 1992;44:147-54.

DAVID M. ROBERTSON, [1] * NICHOLAS CAHIR, [1] HENRY G. BURGER, [1] PAMELA MAMERS, [2] PHILIP I. MCCLOUD, [3] KIM PETTERSSON, [4] and MICHAEL MCGUCKIN [5]

[1] Prince Henry's Institute of Medical Research, Clayton, Victoria 3168, Australia.

Departments of [2] Obstetrics and Gynecology and [3] Mathematics, Monash University, Monash Medical Centre, Clayton, Victoria 3168, Australia.

[4] Department of Biotechnology, University of Turku, Turku SF-20500, Finland.

[5] Department of Obstetrics and Gynecology, University of Queensland, Herston, Queensland 4006, Australia.

[6] Nonstandard abbreviations: IFMA, irnmunofluorometric assay; BSA, bovine serum albumin; TSA-BSA, 50 mmol/L Tris-HCl, pH 7.4, 0.154 mol/L NaCl, 1 g/L Na[.sub.3], and 5 g/L BSA; and CVA, canonical variate analysis.

* Author for correspondence. Fax 61 3 9550 6125; e-mail david. robertson@med.monash.edu.au.
Table 1. Geometric means with 5th and 95th percentiles for serum
CA125 and inhibin in the groups studied.

A. Healthy postmenopausal women and postmenopausal women with
ovarian and nonovarian cancers.
 Discrim.
Cancer group n CA125, kilounits/L index (a,b)

Control 165 9.8 (2.3-41.6) 2%
Serous 65 556 (14-21 900) 89%
Mucinous 21 41.1 (2.5-665) 45%
Granulosa cell 11 59 (0.5-6640) 43%
Miscellaneous 57 155 (4.2-5680) 74%
All ovarian cancers 154 188 (3.3-10 600) 75%
Nonovarian (d) 23 148 (2.6-8350) 75%

 Discrim.
Cancer group Inhibin RIA, (c) units/L index (a,c)

Control
Serous 107 (44-258) 33%
Mucinous 282 (27-2980) 71%
Granulosa cell 1750 (220-13 900) 100%
Miscellaneous 122 (20-719) 35%
All ovarian cancers 156 (19-1310) 44%
Nonovarian (d) 97 (36-262) 22%

 Discrim.
Cancer group Inhibin [alpha] C IFMA, ng/L index (a)

Control 46.5 (14-154) 5%
Serous 95 (14-659) 29%
Mucinous 507 (47-5420) 90%
Granulosa cell 1740 (73-41 700) 100%
Miscellaneous 103 (8.4-1270) 35%
All ovarian cancers 152 (9-2560) 45%
Nonovarian (d) 66 (10-437) 17%

B. Miscellaneous ovarian cancer group from Table 1A broken
down into subgroups.
 Discrim.
Cancer group n CA125, kilounits/L index

Clear cell 10 79 (6-1140) 80%
Endometrioid 19 287 (4.0-23 600) 79%
MMT 5 439 (15-13 000) 100%
Undifferentiated 13 151 (3.0-8020) 69%
Other (e) 10 58 (3.9-1360) 50%

 Discrim.
Cancer group Inhibin RIA, units/L index

Clear cell 88 (34-227) 20%
Endometrioid 109 (26-462) 42%
MMT 99 (34-287) 40%
Undifferentiated 115 (13-1060) 23%
Other (e) 236 (12-4670) 50%

 Discrim.
Cancer group Inhibin [alpha] C IFMA, ng/L index

Clear cell 108 (10-1130) 40%
Endometrioid 124 (18-878) 42%
MMT 102 (18-592) 40%
Undifferentiated 96 (3-2943) 31%
Other (e) 78 (1-4983) 20%

(a) Discrimination between ovarian cancer and control groups was
based on the percentage of values detected above the 95th
percentile of the corresponding healthy postmenopausal control group.

(b) Discrim., discrimination.

(c) Discrimination cutoff of 122 units/L, defined previously (5).

(d) Nonovarian cancers consists of three endometrial, three colon,
four fallopian tube, three intra-abdominal (site unspecified), four
peritoneal (primary site unknown), one stomach, one kidney, and one
breast cancer, and three lymphomas.

(e) Sertoli-Leydig cell, n = 2; Krukenberg, n = 1; pseudomyxoma,
n = 1; mixed cystadenocarcinoma, n = 2; carcinosarcoma of the ovary,
n = 1; mixed epithelial, n = 2; metastatic carcinoma, n = 1.

Table 2. Areas under ROC curves ([+ or -] SE) between assays and
combination of assays (a) for the major ovarian cancer groups.

Cancer group CA125 [alpha] C IFMA

Serous 0.965 [+ or -] 0.0166 (b) 0.745 [+ or -] 0.0385 (c)
Mucinous 0.830 [+ or -] 0.0575 (b) 0.994 [+ or -] 0.0048 (c)
Granulosa cell 0.845 [+ or -] 0.0709 (b) 1.000 [+ or -] 0.0000 (c)
Miscellaneous 0.933 [+ or -] 0.0208 (b) 0.712 [+ or -] 0.0413 (c)
All ovarian 0.929 [+ or -] 0.0150 (c) 0.778 [+ or -] 0.0263 (b)
 tumors

 CA125 1 aC IFMA

Cancer group CVA multiROC

Serous 0.968 [+ or -] 0.0158 (b) 0.966 [+ or -] 0.0176 (b)
Mucinous 0.997 [+ or -] 0.0026 (c) 0.995 [+ or -] 0.0043 (c)
Granulosa cell 1.000 [+ or -] 0.0000 (c) 1.000 [+ or -] 0.0000 (c)
Miscellaneous 0.944 [+ or -] 0.0203 (b) 0.908 [+ or -] 0.0300 (c)
All ovarian 0.960 [+ or -] 0.0115 (d) 0.944 [+ or -]
 tumors 0.0146 (c,d)

(a) The assay combinations were determined by CVA and multiROC
analysis.

(b-d) Comparisons between groups were determined by the method of
Hanley and McNeil (17). Differences between groups within each row
are designated by the superscripts b, c, and d where different
superscripts b vs c, b vs d, and c vs d show significant differences
(P , 0.05), and the same letters, nonsignificant differences.

Table 3. Percentage of cancers detected at 95% and 90% specificity
for the various assays and combination of assays. (a)

 CA125 + [alpha] C
 IFMA

 CA125 [alpha] C IFMA CVA MultiROC

95% specificity 28.5 units/L 129 ng/L
 Serous 92% 33% 94% 92%
 Mucinous 50% 95% 100% 95%
 Granulosa cell 57% 100% 100% 100%
 Miscellaneous 79% 46% 81% 82%
 All ovarian 80% 50% 89% 90%
 tumors

90% specificity 23.3 units/L 98.2 ng/L
 Serous 94% 44% 94% 94%
 Mucinous 65% 100% 100% 100%
 Granulosa cell 57% 100% 100% 100%
 Miscellaneous 82% 47% 86% 88%
 All ovarian 84% 56% 91% 93%
 tumors

(a) Data are derived from the corresponding ROC curves. The
corresponding analyte concentrations at 95% and 90% specificity
are also included.

Table 4. Regression coefficients between aC IFMA and
inhibin RIA for serum inhibin from healthy postmenopausal
women and postmenopausal women with ovarian and
nonovarian tumors.

 Regression Correlation
Cancer group equation (a) coefficient n

Serous y = 0.14x + 1.8 r = 0.31 (b) 64
Mucinous y = 0.82x + 0.22 r = 0.83 (c) 21
Granulosa cell y = 0.41x + 1.9 r = 0.63 (b) 11
Miscellaneous y = 0.33x + 1.4 r = 0.46 (c) 57
All ovarian tumors y = 0.81x + 0.41 r = 0.78 (c) 153
Nonovarian tumors y = 0.26x + 1.5 r = 0.50 (b) 23

(a) Abscissa (x), [alpha] C IFMA; ordinate (y), RIA. Data were
log-transformed prior to analysis.

(b) P <0.05.

(c) P <0.001.
COPYRIGHT 1999 American Association for Clinical Chemistry, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1999 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Enzymes and Protein Markers
Author:Robertson, David M.; Cahir, Nicholas; Burger, Henry G.; Mamers, Pamela; McCloud, Philip I.; Petterss
Publication:Clinical Chemistry
Date:May 1, 1999
Words:5676
Previous Article:Anti-free prostate-specific antigen monoclonal antibody epitopes defined by mimotopes and molecular modeling.
Next Article:Measurement of plasma renin activity with use of HPLC-electrospray-tandem mass spectrometry.
Topics:


Related Articles
Bcl-2 urine test for ovarian ca gets good results in pilot study.
Ovarian cancer continues to defy early detection.
Multiplexed proximity ligation assays to profile putative plasma biomarkers relevant to pancreatic and ovarian cancer.
A novel, high-throughput workflow for discovery and identification of serum carrier protein-bound peptide biomarker candidates in ovarian cancer...
Predicting response of ovarian cancer to paclitaxel treatment based on trend analysis of serum CA125.
Unfavorable prognostic value of human kallikrein 7 quantified by ELISA in ovarian cancer cytosols.
HEALTHLINX CONFIRMS COLLABORATORS FOR OVPLEX OVARIAN CANCER STUDY.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters