An improved statistical approach: can it clarify the role of new prognostic factors for breast cancer?

Recently, there has been a proliferation of new biomarkers, some of which may lead to an improved prognostic index or may influence treatment selection. However, there are methodological and statistical issues that require attention in assessing the role and use of these prognostic factors. Between 1977 and 1986, 1097 primary breast cancer patients were accrued for multidisciplinary research at the Henrietta Banting Breast Centre, Women's College Hospital; follow-up to 1990 is complete for 96% of the patients. Data for these patients are used here to illustrate strategies: (1) for the comparison of results from diverse assessments of biomarkers; (2) for the improved comparability of inter-laboratory results; (3) for the examination of the results from monoclonal or polyclonal antibody assays for possible clinically relevant bimodality; (4) for good statistical resolution of overlapping distributions; (5) that involve the use of quantitative values for prognostic factors whenever possible; and (6) for improved multivariate analyses. Good data handling and analyses may enable more accurate and rapid assessment of new prognostic factors, thereby expediting and improving their clinical application.

An investigation of cut-points for primary breast cancer oestrogen and progesterone receptor assays.

Oestrogen and progesterone receptor (ER and PgR) assay values are frequently used in medical decision-making for breast cancer patients. We have proposed statistical standardization of receptor assay values to improve inter-laboratory comparability, and now report the use of standardized log units (SLU) to investigate the effects of ER and PgR cut-points on time to first recurrence outside the breast (DFS). Between 1980 and 1986, there were 678 primary breast cancer patients treated at the Henrietta Banting Breast Centre (HBBC). The effects of ER and PgR cut-points were examined with multivariate analyses considering the variables: age, tumour size, nodal status, weight and adjuvant treatment. We considered receptor assay cut-points ranging from - 1.0 to + 1.0 SLU (ER between 7 and 166 fmol/mg protein; PgR between 7 and 181 fmol/mg protein). PgR was included in the multivariate prognostic models more often than ER, although patients had a better prognosis with both larger ER and PgR values. There was no best cut-point for ER or PgR, and there was strong evidence that ER and PgR should be considered as continuous rather than dichotomous (negative, positive) variables. Patient prognosis should also be more comparable with SLU.

The Henrietta Banting Breast Centre database: a model for clinical research utilizing a hospital-based inception cohort.

The cohort study design has been used successfully in clinical cancer research. Cohorts, however, are valuable only if they produce results which are valid and generalizable. Some hospital-based inception cohorts satisfy both these requirements and may thus be useful research tools. The development of one such hospital-based cohort, the Henrietta Banting Breast Centre database, is described. This cohort is composed of 1097 women diagnosed with primary breast cancer at Women's College Hospital, Toronto, from January 1977 through December 1986. Details of diagnostic procedures, pathology, treatment, dates and sites of recurrence, and date of death are available on 96% of women. By comparison with published series and with the Ontario Cancer Registry, we have demonstrated validity and generalizability. A major advantage is the ready availability of paraffin tissue blocks on virtually all cases, facilitating analyses of the prognostic importance of specific biologic variables and immunocytochemical hormone assays. Other completed studies and future uses of the cohort are described.

The predictive value of ERICA in breast cancer recurrence. A univariate and multivariate analysis.

In breast cancer, primary tumor size (T), the number of lymph node metastases (#N), the biochemical estrogen (ER), and progesterone (PGR) receptor status have all been important prognostic variables. We evaluated the significance of the immunocytochemical measurement of estrogen receptors suing the ERICA method. To determine the relative prognostic value of these variables T, #N, ER, PGR, ERICA and adjuvant treatment, (ADJ), univariate and multivariate analyses of disease-free survival (DFS) were performed for 154 primary breast cancer patients who were diagnosed in 1985 to 1986 at Women's College Hospital and followed prospectively. We analyzed ERICA results using different classification systems, and assessed clinical cut points for the univariate and multivariate context. The variables consistently included in the best Cox stepwise regression are T, (p < 0.01), ADJ (p < 0.01), #N (p < 0.01), and ERICA (p < 0.01). There was weaker evidence of an association between DFS and the biochemically determined ER; ER was not included in the model with a cut point at 10 fmol mg of protein. This illustrates the value of the ERICA method in predicting outcome, and suggests the need to consider ERICA values for clinical decision making.

The standardization of estrogen receptors.

Tumour estrogen receptor (ER) status may determine the medical treatment of a patient with breast cancer; yet inter-laboratory results can vary markedly, particularly when absolute cut-offs in fmol/mg cytosol protein are used. The use of standardized log units is proposed to permit greater inter-laboratory comparability. We have assessed the biochemical ER values using the dextran-coated charcoal method with three data sets, two quality control (QC) sets for Ontario laboratories and a data set with values for 184 primary breast cancer patients seen at Women's College Hospital (WCH) between 1985 and 1986. The distributions for all the raw data were skewed toward the lower end of the range; a log transformation improved the symmetry of the distributions. There was marked inter-laboratory variation in the QC data, and standardized log units greatly reduced this variability. The WCH data had similar differentiation by tumour size and nodal status with both the raw data and standardized log units. However, standardized log units provided more consistent evidence of an association between ER and immunohistochemical ERICA. The standardized log units provide quantitative receptor values suitable for multi-centre research, for future work with clinical outcomes, and for the daily management of patients.

Evidence for bimodal distribution of breast carcinoma ER and PgR values quantitated by enzyme immunoassay.

Breast carcinoma oestrogen receptor (ER) and progesterone receptor (PgR) values obtained by radioligand binding assays have commonly been observed to have approximate log-normal distributions. We examined the distribution of log-transformed receptor values obtained by enzyme immunoassay for 5468 primary breast carcinomas in five Ontario laboratories. In each laboratory, it was found that the frequency histograms for the log transformed receptor values were not unimodal, and generally were suggestive of bimodality. This was not affected by stratification by age or inferred menopausal status (< or = 49, > or = 50 years), and could not be explained by kit characteristics. However, the low point in the distribution varied from 5 to 63 fmol/mg cytosol protein, depending on the receptor, patient age and laboratory. The tendency towards biomodality was more distinct for ER than for PgR. It remains to be determined whether the low points on the frequency histograms have clinical relevance for discriminating between hormone-sensitive and hormone-insensitive tumours.

A comparison of all-subset Cox and accelerated failure time models with Cox step-wise regression for node-positive breast cancer.

Clinical studies usually employ Cox step-wise regression for multivariate investigations of prognostic factors. However, commercial packages now allow the consideration of accelerated failure time models (exponential, Weibull, log logistic, and log normal), if the underlying Cox assumption of proportional hazards is inappropriate. All-subset regressions are feasible for all these models. We studied a group of 378 node positive primary breast cancer patients accrued at the Henrietta Banting Breast Centre of Women's College Hospital, University of Toronto, between January 1, 1977, and December 31, 1986. 85% of these patients had complete prognostic factor data for multivariate analysis, and 96% of the patients were followed to 1990. There was evidence of marked departures from the proportional hazards assumption with two prognostic factors, number of positive nodes and adjuvant systemic therapy. The data strongly supported the log normal model. The all-subset regressions indicated that three models were similarly good. The variables 1) number of positive nodes, 2) tumour size, and 3) adjuvant systemic therapy were included in all three models along with one of three biochemical receptor variables 1) ER, 2) combined receptor (ER- PgR-; ER+PgR-; ER- PgR+; ER+PgR+; or 3) PgR. Better multivariate modeling was achieved by using quantitative prognostic factors, a check for appropriate underlying model-type, and all-subset variable selection. All-subset regressions should be considered for routine use with the many new prognostic factors currently under evaluation; it is very possible that there may not be a single model that is substantially better than others with the same number of variables.

Suppression of the growth of the androgen-insensitive R3327 HI rat prostatic carcinoma by combined estrogen and antiprogestin treatment.

The antiprogestin RU486 has been shown to inhibit the growth of a number of tumor cell lines and solid tumors which contain significant concentrations of progesterone receptor (PgR). It has been suggested that growth suppression may be due to a PgR-mediated cytotoxic effect. The R3327 HI prostatic carcinoma of the rat is considered to be a model for human prostatic carcinoma which has become resistant to androgen deprivation therapy. Since it is possible to induce high concentrations of PgR in this tumor with estrogen, it was of interest to investigate the possibility that RU486 could suppress its growth. Growth was assessed by tumor diameter, [3H]thymidine uptake and histopathological appearance after 2 or 8 weeks treatment with RU486 alone, diethylstilbestrol (DES) alone, and combined RU + DES treatment as compared with control animals. Tumor growth was not affected significantly by DES treatment alone. RU486 treatment alone suppressed PgR content and resulted in only insignificant inhibition of growth. However, when significant PgR concentrations were maintained by combined treatment with DES, RU486 significantly suppressed tumor growth (0.01 less than P less than 0.05 vs controls). This was accompanied by atrophy of the glandular epithelium. The results support the hypothesis that growth suppression may be brought about by a PgR-mediated mechanism. The data suggest that it may be possible to treat androgen-insensitive prostatic carcinoma by a new form of hormonal treatment.

Quantitation of cytosolic and nuclear estrogen and progesterone receptor in benign, untreated, and treated malignant human prostatic tissue by radioligand binding and enzyme-immunoassays.

Estrogen receptor (ER) and progesterone receptor (PgR) were quantitated in benign and malignant human prostatic tissue by using radioligand binding assays (RBA) and enzyme-immunoassays (EIA). Using a hydroxylapatite exchange method for ER, little or no nuclear ER (ERN) could be detected, but with the EIA both cytosolic (ERC) and ERN were detected in almost all specimens, although in meager concentrations. Tissue from patients with carcinoma had significantly higher ERC concentrations than tissue from patients with benign disease. Specimens from estrogen-treated patients had significantly lower ERC:ERN ratios than those from untreated patients. Progesterone receptor was detected in virtually all specimens by both methods, at concentrations higher than those of ER. Carcinoma tissue with a high malignant involvement contained significantly less PgR than benign tissue. Using either method, the highest concentrations of PgR were observed in tissue from carcinoma patients treated with estrogen. Overall, the data suggest that although human prostatic tissue contains only modest amounts of ER, this is active in that treatment with estrogen promotes association of this receptor with the nuclear fraction and increases PgR content.

Immunohistochemical localization of progesterone receptor in benign and malignant human prostate.

A monoclonal antibody to progesterone receptor, KD68, was used to localize this receptor protein in 31 surgical specimens of benign and malignant human prostate. Progesterone receptor was detected almost exclusively in stromal cells. The most striking finding was the periacinar arrangement of stained cells in some specimens of glandular BPH, sometimes associated with close apposition of stained cells to the basement membrane of the acinar epithelium. In both benign and malignant specimens scattered stained cells were observed in some areas of fibromuscular stroma. Except for occasional stained epithelial cells in one benign and one estrogen-treated malignant specimen, both benign and malignant epithelium were negative.

Changes in tumor characteristics during progression of the R3327 HI experimental prostatic carcinoma.

The R3327 HI experimental prostatic carcinoma was serially transplanted through six generations of castrated host rats to examine changes in the capacity to synthesize progesterone receptor (PgR) in response to diethylstilbestrol (DES) stimulation during progression from a well-differentiated state containing a significant stromal component to a poorly differentiated state with virtually no stroma. During progression, changes in growth rate and histopathology occurred in stepwise fashion, the most marked changes being observed between the first and second and between the fourth and fifth generations. The capacity for PgR synthesis in response to DES treatment fell progressively, but did not reach statistical significance during the first four generations. By the sixth generation, the stromal component had virtually disappeared, and no estrogen receptor (ER) or PgR was detectable. During the course of the investigation, a monoclonal antibody that reacts with rat PgR became available, and immunohistochemistry with this antibody confirmed that DES-induced PgR was present in stromal cells. We conclude that this work supports the hypothesis that prostatic stroma is a target tissue for estrogen and that this model may be useful for the investigation of other events associated with progression.

Comparative evaluation of ER-ICA and enzyme immunoassay for the quantitation of estrogen receptors in breast cancer.

Evaluation of estrogen receptors (ERs) is essential for the treatment and prognosis of human breast cancer. The authors have undertaken a study of 225 primary breast cancers to compare the receptor values using immunohistochemistry and enzyme immunoassay (EIA) techniques. The results are compared with the biochemical assay with the use of the dextran-coated charcoal (DCC) method done on the same specimens. The overall concordance between ER-ICA and DCC was 77%. Ninety-seven percent concordance was observed for specimens with a positive ER-ICA score (greater than 70 and the DCC more than 10 fmol/mg protein). Ninety-two percent of tumors had negative results by ER-ICA when the DCC was less than 10 fmol/mg protein. When the ER-ICA score was plotted against the DCC or EIA value, a significant correlation was obtained with a P value of less than 0.001. The correlation between the ER-ICA score and the total ER content (cytosol and nuclear) assayed by EIA was similar to comparison with cytosol alone. The authors conclude that the ER-ICA can be useful in assessing ERs in breast especially in hypocellular or small tumors and in assessing the degree of heterogeneity in a given tumor. The immunohistochemical and biochemical assay methods are complementary and provide greatly needed information for the management of breast cancer.

Basal and estrogen-stimulated hormone receptor profiles in four R3327 rat prostatic carcinoma sublines in relation to histopathology and androgen sensitivity.

Four sublines (H, HI, G, and AT) of the R3327 (Dunning) rat prostatic carcinoma have different androgen sensitivities and histopathological characteristics. In order to investigate whether these characteristics were associated with differences in the hormone receptor profile and its response to estrogen, we carried out Scatchard analysis on the cytosolic (C) and high-salt nuclear-associated (N) androgen (AR), estrogen (ER), and progesterone (PgR) receptor in each line, carried in control and diethylstilbestrol (DES)-treated animals. In the H line (androgen sensitive, well differentiated) DES treatment resulted in significant increases in total cellular AR and ER, in redistribution of both receptors between the C and N fractions, and in a marked increased of PgR (greater than 10-fold). The hormone receptor profile and its response to DES was similar in the HI line (androgen insensitive, well differentiated), except that total cellular ER was not increased after treatment. The G line (androgen sensitive, poorly differentiated) contained higher basal concentrations of AR and PgR than the H line, but the concentrations were not increased by DES treatment, although treatment promoted association of ER with the nuclear fraction. The AT line (androgen insensitive, anaplastic) contained no ER and negligible PgR, but AR was present, although in lower concentrations than in the other lines. Diethylstilbestrol treatment had no effect on the concentration, although redistribution of AR between C and N fractions did occur. Some characteristics of the AR in the AT line differed qualitatively from that in the H line, but injection of testosterone into castrated animals bearing the AT tumor promoted association of AR with the nuclear fraction, indicating normal activation. The data suggest that the ability of DES treatment to increase AR and PgR concentrations is associated with differentiation and/or the presence of stroma and that it is unrelated to androgen sensitivity.

Use of an enzymeimmunoassay (EIA) for quantitation of cytosolic and nuclear estrogen receptor, and correlation with progesterone receptor in human breast cancer.

We have adapted a commercially available enzyme immunoassay for ER (ER-EIA) for use with nuclear extracts of breast carcinoma specimens, and have examined the data obtained in relation to the results from cytosolic ER-EIA and radioligand (DCC) assays and from DCC assays for PgR. In a series of 139 carcinoma specimens, we observed a very significant correlation between the cytosolic ER concentration as measured by the EIA and DCC methods, and also between the log10 of the concentration of ER in the cytosolic and nuclear fractions assayed by the ER-EIA method. The correlation between nuclear ER and cytosolic PgR was also highly significant, but not as close as for cytosolic ER. If 130 fmol ER/mg DNA was used as a "cut-off" point to discriminate between specimens "positive" and "negative" for ERN, there was 87% concordance in receptor status between ERN and cytosolic ER, and 79% concordance between ERN and cytosolic PgR. Forty-one percent of specimens were positive or borderline for both ERN and cytosolic PgR, and it is suggested that this receptor combination may be a valuable predictive factor in prognosis and response to hormonal treatment.

Age as a confounding factor in the association of mammographic dysplasia and estrogen receptor concentration in breast cancer.

We have examined the association between hormone receptor concentration in primary breast cancer and the mammographic pattern of the breast in which the cancer arose. A significant association was found between the concentration of estrogen receptor and the proportion of the breast volume occupied by the radiological signs of dysplasia. Both estrogen receptor concentration and dysplasia were found to be strongly associated with age. Estrogen receptor concentration rose with increasing age, while the age of patients with extensive dysplasia was substantially less than that of patients with no dysplasia. After taking age into account, no association remained between estrogen receptor concentration and mammographic dysplasia. Age is therefore a confounding factor in this association.

Estrogen and progesterone receptor content of primary and secondary breast carcinoma: influence of time and treatment.

ER and PgR concentrations were assayed in primary and secondary breast carcinoma specimens from patients classified into 3 groups: (1) both specimens excised on the same occasion (61 patients); (2) specimens obtained on separate occasions with no intervening treatment (43 patients); (3) specimens obtained on separate occasions with intervening chemotherapy and/or irradiation (25 patients). There were highly significant linear correlations (P less than 0.001) between the concentrations of ER (expressed as log10) in primary and secondary specimens in all groups. The relationship between PgR concentrations in primary and secondary specimens in groups 1 and 2 was highly significant, although there appeared to be a greater tendency for loss of PgR in sequential, than in simultaneous secondary biopsies. When expressed in terms of hormone receptor status (HRS), the same rate of discordance was observed in groups 1 and 2 (30% when concentrations were expressed in terms of cytosol protein). In group 1 the major cause of discordance was the occurrence of receptor +ve secondaries in association with receptor -ve primaries, possibly because of the high cellularity of many involved axillary nodes. In group 2, the major cause of discordance was the occurrence of receptor -ve secondaries derived from receptor +ve primaries. In both groups discordance in PgR status was more frequent than in ER status. In group 3, overall discordance in HRS was 24% and was due equally to ER and PgR; however, the high concordance rate for PgR was probably due to the fact that the tumours were initially PgR -ve, and the secondaries were also -ve. These results confirm that ER content tends to be stable, even after long periods of time and the administration of chemotherapy and/or irradiation. Progesterone receptor content is much less stable, and may decrease during quite short time intervals even in the absence of treatment.

Regulation of estrogen and progestin receptor concentrations in an experimental rat prostatic carcinoma by estrogen, antiestrogen, and progesterone.

In order to assess prostatic tissue as a target for receptor-mediated estrogen action, we have examined the regulation of estrogen (ER) and progestin receptors (PgR) by estrogen, antiestrogen, and progesterone in cytosolic and nuclear fractions of the R3327H (Dunning) prostatic adenocarcinoma of the rat. Twenty micrograms diethylstilbestrol (DES) with or without 800 micrograms tamoxifen (Tam) were injected s.c. in oil 5 times weekly for 2 weeks. Controls were given oil only. Estrogen receptor assays were carried out using [3H]estradiol and a hydroxylapatite exchange method. Progestin receptors were assayed using [3H]R5020 and dextran-coated charcoal to separate free and bound steroid. All binding data were evaluated by using Scatchard analysis. Treatment with DES depleted cytosolic ER, promoted association of ER with the nuclear fraction, and concomitantly increased PgR concentrations in amounts proportional to nuclear ER. Treatment with Tam alone resulted in higher nuclear ER concentrations than treatment with DES, but induced only one-fifth the amount of PgR. Treatment with DES plus Tam resulted in similar nuclear ER concentrations as with Tam alone, but PgR concentrations were intermediate between those observed with DES alone and Tam alone. Thus Tam exhibited both estrogenic and antiestrogenic properties. In this experiment, the same cytosolic and nuclear extracts were also assayed for ER by using monoclonal antibodies to the receptor in an enzyme immunoassay. No significant differences were observed between the results obtained by the radioligand and enzyme immunoassay methods in the cytosol and nuclear fractions from the control and DES-treated tumors. However in both Tam-treated groups, the ER values obtained by the enzyme immunoassay method were significantly higher than those obtained by the radioligand method in both cytosolic and nuclear fractions. This confirms the observations made by others in female target organs, that monoclonal antibody to ER reacts differently with the Tam-bound ER complex than with the estradiol-bound ER complex. In a separate experiment, administration of progesterone with DES decreased the concentration of nuclear ER to less than one-half that observed after administration of DES alone, with proportional decreases in both cytosolic and nuclear PgR. All these observations indicate that the control of ER and PgR concentrations in this prostatic tumor is identical to that observed in female rat target organs. Use of an immunohistochemical method for the detection of ER in frozen sections indicated that the receptor was localized in the glandular epithelium in both control and DES-treated tumors.

Role and mechanism of action of tamoxifen in premenopausal women with metastatic breast carcinoma.

Tamoxifen was evaluated as initial hormone therapy for metastatic breast cancer in 85 premenopausal patients. Tamoxifen responders continued on tamoxifen, while tamoxifen failures and initial responders who later progressed were to receive ovarian ablation next. Of 74 evaluable patients, 5 had complete responses (CR) and 15 had partial responses (PR) while 12 remained stable (ST), giving response rates of 27% (CR + PR) or 43% (CR + PR + ST). Of the 23 patients who initially responded (CR + PR + ST) to tamoxifen but then progressed and received ovarian ablation alone, 15 are assessable. Nine (60%) responded (CR + PR + ST) to ovarian ablation. Sixteen patients who failed tamoxifen had ovarian ablation alone, and of 14 assessable patients 2 had ST while 12 progressed. Thus response to tamoxifen strongly predicted response to ovarian ablation (P = 0.021). Serial follicle stimulating hormone, prolactin, and estradiol levels suggested that tamoxifen does not act by induction of a "medical ovariectomy" or by alteration of prolactin levels in premenopausal patients.

Quantitative relationships between cytosol and nuclear estrogen and progesterone receptors in the R3327 prostatic carcinoma of rats treated with diethylstilbestrol.

Estrogen and progesterone receptor (ER, PgR) were quantitated by Scatchard analysis in cytosolic and nuclear extracts of R3327H prostatic carcinomas from rats injected with 100 micrograms diethylstilbestrol or with vehicle for 1-3 weeks. Total tissue ER concentrations were less than 80 fm/mg DNA in control tumors and were not significantly higher in treated tumors: however, in the latter, a significantly higher proportion (73 +/- 16%) was associated with the nuclear fraction than in the control tumors (less than 33%). Mean PgR concentrations were more than seven-fold higher in treated than in control tumors, and approximately 25% of the total was associated with the nuclear fraction in both treated and control tumors, under the assay conditions used. There was a strong linear correlation between nuclear ER and PgR concentrations. These data indicate that although ER concentration in these tumors was low compared with that in female target organs, it was functional in its ability to associate with the nuclear fraction and induce synthesis of an estrogen-induced protein (PgR) in proportionate amounts.

Inter-laboratory quality control of estrogen and progesterone receptor assays in breast cancer tissue using lyophilised cytosols.

In 1981 a quality control (QC) program for estrogen and progesterone receptor assays was organized among six laboratories in Ontario, Canada. Twenty-three vials of lyophilised cytosol prepared from human breast tumor tissues were analysed by each laboratory over a two-year period. Samples of each batch of QC material were analysed at least twice: either in the same batch or on separate occasions. The present study demonstrates the stability of the QC material, defines the relative accuracy of the receptor assays, and provides estimates of within-batch and between-batch precision of the receptor assays.