Environmental factors and breast cancer.

This review summarizes the results of studies on the effects of environment on breast cancer risk. As known risk factors such as reproductive life, inheritance, and socioeconomic status are estimated to explain only about half of the breast cancer cases, it has been thought that environmental factors could also be related to the risk of this disease. It is known that ionizing radiation is an environmental risk factor increasing the risk of breast cancer. The data of experimental studies show that some organochlorines could be associated with breast cancer risk although the data from epidemiological studies are not consistent due to the difficulties to assess exposure and other risk factors. Recent experimental studies show that cadmium is an environmental factor that mimics the effects of estradiol in estrogen-responsive breast cancer cell lines while solar radiation possibly decreases the risk due to protective effect of vitamin D. The data on the effect of electromagnetic fields are not consistent. Although evidence about the effect of environmental factors on the risk of breast cancer is not convincing, some of these factors together with inheritance, reproductive life, and age at exposure could be associated with an increased risk of the disease.

SEVERAL ATTEMPTS have been made to diagnose cancer using serological techniques. More recently, the use of simultaneous assays has enabled the detection of a higher proportion of patients (Franchimont et al., 1976;Coombes et al., 1980). In gynaecological cancer, a number ofworkers have looked for raised levels of various cancer-related substances which may correlate with tumour activity, including carcinoembryonic antigen (CEA), o-foetoprotein (AFP), f subunit of human chorionic gonadotropin (hCG), human placental lactogen, isoenzymes, etc. (Seppala et al., 1975;Fishman, et al., 1975;Lin et al., 1975;Samaan et al., 1976;Rutanen & Seppala, 1978;Khoo et al., 1977Khoo et al., , 1979a. In most of these studies, the predictive value of these tests is not clearly established. Moreover, in studies of this nature the use of a control group of patients is critical, since tumour markers such as AFP and hCG may be under the influence of hormones which may precede tumour development and may be active long after removal of the tumour. In this study we examine the value of the oncofoetal antigens (namely, CEA, AFP, and hCG) in patients with cancer of the ovary, squamous-cell carcinoma of the cervix, and adenocarcinoma of the cervix, and compare these findings with those in cancer patients who have been free of tumour for at least 1 year, as well as with age-matched non-tumour-bearing patients.

Patients
This study comprised 84 patients with cancer of the ovary or cervix who were examined pre-operatively. There were 30 patients with cancer of the ovary, the histological diagnosis being serous adenocarcinoma (11), endometrioid adenocarcinoma (6), mucinous adenocarcinoma (4), mixed tumour (4) and others (5). Most patients were in were again washed and counted. The sensitivity of the assay (minimum concentration ith 3 in Stage I, 6 in Stage 1, distinguishable from zero) was 1-0 ,ug/l. V (Table I,) AFP.-Serum AFP levels were estimated patients with squamous-cell with a modified RIA kit supplied by Amerie cervix, of whom 16 had sham (U.K.). Serum (100 ,ul) or standards eratinizing carcinoma. Most (0-200 ,g/l) were incubated with 100 ul of wereiinStagcar(16)mand Mot antibody for 6 h at 37°C, then 100 pi of were in Stage I (16) and II 125I-labelled AFP added and incubation con-)up of 21 patients were diagtinued at room temperature (25°C) for another ocarcinoma of the cervix, 12 h. The bound fraction was separated by I (14). The histological diagadding 1000 pl of polyethylene glycol (PEG) follows: well differentiated at 200 g/l, w/v, and the precipitates collected (6), adenosquamous carcinby centrifugation at 3000 g for 20 min and differentiated adenocarcin-counted for 1 mm. Sensitivity of the assay ell carcinoma (1), undifferwas 0-1 ,ug/l. Intraand inter-assay copcienoma (2) ohadprerso(3 efficients of variation at 15 ,ug/l were 6.2% patients who had previously and 4.2% respectively. r cancer of the ovary and hCG.-Serum hCG levels were estimated by t the time of study had been RIA kit supplied by Mallinckrodt (St Louis, cr for at least 1 year, were Missouri, U.S.A.). Serum (100 pl) or standard control group (tumour-free (0-100 i.u./1) were incubated with 100 PI of hCG antiserum and 100 ul of 1251-labelled our control group, 24 age-hCG for 18 h at room temperature (25TC). The os with non-malignant gynaebound fraction was separated by adding 2 ml itons were studied The age of PEG (200 g/l, w/v) and the precipitate atients in the various groups collected by centrifugation at 3000 g for 20 e min. Cross reactivity of the antisera with LH, FSH, and TSH was not significant, being ye distribution in patients 0 11, 0 -11, and 0 80% respectively. Inter-and studied intra-assay coefficients of variation at 9-5 Age (years) i.u./l were 7-6% and 11.6% respectively. The Age (year) Asensitivity of the assay was 0-025 i.u./l. The values of oncofoetal antigens in the various cancer patients and control groups are shown in Table III. There were marked variations in all tests, rendering mean values of little significance. Of more relevance is the proportion of patients in each group with high values.
CEA levels above 2-5 ,ug/l were found in 54% of all cancer-bearing patients, as well as in 39% of tumour-free cancer patients and 38% of non-cancer controls (Table   IV). When a cut-off point of 10 ,ug/l was chosen, only 8% of cancer patients were positive, compared to 2% of tumour-free patients previously treated for cancer, while none of the non-cancer control group were positive. Patients with adenocarcinoma of the cervix had a higher frequency of high values. AFP was > 5 ,ug/l in 10% of cancer patients and in 4% of non-cancer controls.
When the cut-off point was increased to 10 ptg/l, 7% of the cancer patients had high values, while none of the tumour-free patients or the non-cancer control group had high values. The highest percentage of patients with high AFP was in the group with adenocarcinoma of the cervix (19%). hCG was > 3 i.u./l in 20% of cancer patients and in 27% of tumour-free patients. Values of hCG > 10 i.u./l were found in 5% of all cancer patients and in none of the cancer-free patients or noncancer controls.
The proportion of patients with one or more of these tests positive is shown in Table V. When a low cut-off point was used (i.e. CEA> 2-5 ,ug/l, AFP> 5 ,g/l, hCG> 3 i.u./l, 67% of cancer patients were found to be positive, but this was not significantly different from tumourfree cancer patients (54%) or the nontumour group (42%). Likewise, when the cut-off point was slightly higher (viz. CEA > 5 P,g/l, AFP > 10 jug/l, and hCG > 10 i.u./l), the proportion of patients with one or more positive tests was 15% for the tumour-free cancer patients and 8% in the non-tumour group, compared to 30% in the cancer patients. When the cut-off point was raised even further (CEA>10 ,ug/l, AFP>10 ,ug/l, and hCG > 10 i.u./l), none of the control group had high levels, compared to 17% of the cancer-bearing group or 2% of the tumourfree cancer patients. Patients with adenocarcinoma of the cervix were positive for one or more tests more frequently (33%0) than other cancer groups.

DISCUSSION
The value of the detection of CEA in gynaecological cancer has been subjected to several studies. When a level of 2-5 ,ug/l is taken as the cut-off point, an unacceptably high false-positive rate is detected in normal controls, e.g. 11% (Van Nagell et al., 1975), 18% (Donaldson et al., 1980) and 10% (Di Saia et al., 1977). When a CEA level of 5,Lg/l is taken as the cut-off point, the proportion of patients with high levels varies from 63% in cancer of the ovary (Khoo et al., 1977(Khoo et al., , 1979aSarjadi et al., 1980), 31% of cases of cancer of the corpus, 36% of patients with cancer of the cervix, and 36% of cancer of the ovary, compared to 0% in controls. Rutanen et al. (1978), on the other hand, found an incidence of only 9-8% in patients with gynaecological cancer, with a maximal incidence in ovarian cancer (20%); squamous-cell carcinoma had 10%, adenocarcinoma of the cervix 19%, and endometrial carcinoma 7%. In our study we find that, when a CEA value of 2-5 ptg/l is taken as cut-off, 38% of controls have high values, compared to 54% of cancer patients. This figure is quite unacceptable. When, however, a cut-off point of 10 ug/l is taken, none of the control patients have a high value, compared to 8 % of cancer patients. 14% of adenocarcinoma of the cervix patients had high CEA lev-els, compared to 3% with squamous-cell carcinoma. Moreover, in a previous study we showed that in none of 36 patients with cancer of the endometrium was CEA> 5 ,ug/l (Cauchi et al., 1980). This is in agreement with the findings of Franchimont et al. (1976), Hansen et al. (1974), and Stone et al. (1977, who emphasize the importance of taking CEA> 10 /xg/l as the cut-off level. AFP has also been investigated as a possible tumour marker in gynaecological cancer. Khoo et al. (1977) found that 17% of 108 patients with cancer of the ovary had AFP levels > 25 ,ug/l. In germ-cell tumours the level was usually > 200 ,ug/l. Donaldson et al. (1980) found values of AFP > 20 ,ug/l in 52% of invasive cancers (endometrial cancer 50%, ovarian cancer 57%, vulval cancer 43%, cervical cancer 53%). However, the finding that 22% of control patients also have high AFP cannot be readily explained. These authors conclude that AFP levels were most often high in large-cell non-keratinizing cancer, as well as germ-cell or stromal tumours of the ovary. Our studies show that none of the control patients or the tumour-free cancer patients had AFP levels > 10 /g/l, whilst 7% of cancer-bearing patients had high values. Human chorionic gonadotropin (hCG) has also been used as a marker of gynaecological cancer. Donaldson et al. (1980) showed that a level of hCG > 5.0 i.u./l was found in 3% of controls and in 22% of invasive gynaecological cancer, the highest values being found in serous cystadenocarcinomas of the ovary and in patients with keratinizing squamous-cell carcinoma of the cervix. Likewise, Carenza etal. (1980) found that 44% of 18 patients with endometrioid cancer and 7/17 (41%) with ovarian cancer had detectable quantities of hCG in their sera, mean hormone levels being 28-4 i.u./l in ovarian cancer and 7-1 i.u./l in endometrioid cancer, and only in 3/34 (8.8%) in benign disease of endometrium or ovary. Although Franchimont et al. (1976) consider level of hCG> 15 ,ug/l to be abnormal, we find that 8% of controls and 27% of tumour-free cancer patients had values >3 i.u./l. However, neither of these groups had levels > 10 i.u./l, whilst 5% of cancer patients had The value of multiple tumour markers to establish the diagnosis of gynaecological cancer depends also on the cut-off point for these markers. A number of workers have used a combination of CEA, AFP, and hCG to detect the presence of gynaecological cancer. Donaldson et al. (1980) found that -85% of gynaecological cancers have elevation of one or more of these cancer markers. However, the cut-off point taken by these authors  Hg/l, hCG 5 0 i.u./l, and AFP 20 ,g/l) produced an unacceptably high level of false positives in control patients (31 %). This is due to the relatively low cut-off point for CEA and the unexplained high proportion of control patients (22%) with AFP> 20 ,ug/l. The reason for this high proportion of AFP positive patients is not clear. Seppala et al. (1975) measured these markers in advanced ovarian cancer and found high CEA in 21% of patients, only one of whom had high AFP and none raised hCG. Our data (Table V) show that, while a low cut-off point for these markers (namely CEA 2-5 ,ug/l, AFP 5 ,6gll, hCG 3 i.u./l) results in an unacceptably high false-positive rate in control patients (42-54%), using a higher cut-off point (CEA 10 pg/l, AFP 10 ,ug/l, hCG 10 i.u./l) produced none of the control patients and only 1/42 of tumour-free cancer patients with one or more positive tests, compared to 17% of cancer:bearing patients. Even higher values (33 %) were found in patients with adenocarcinoma of the cervix.
These studies emphasize the importance of establishing the upper levels of normal, not only for age-matched normal persons but also for tumour-free cancer patients, in view of the fact that factors, including hormone stimulation, might be operating in a cancer patient irrespective of the presence or absence of tumour. The finding that there is a significant correlation between AFP and hCG in patients with adenocarcinoma of the cervix would indicate that production of these hormone markers was the result of the same stimulus, which is not necessarily the tumour itself. Further studies of the relevance of hormone stimulation to high oncofoetal antigen levels are under way.