Characteristics of chromate workers' cancers, chromium lung deposition and precancerous bronchial lesions: an autopsy study.

The characteristics of lung cancers induced by inhaled chromate were studied in 13 consecutive autopsies on male ex-chromate workers. In addition to histopathology, we examined: (1) the relationship between the occurrence of lung cancer and the amount of chromium (Cr) deposited in the lung as determined by atomic absorptiometry and (2) the chronological changes in five precancerous lung lesions followed by bronchoscopy till death. Twenty-one cancers were identified, including 16 lung tumours observed either during follow-up or at autopsy. Of these 16 tumours, 13 were found in six subjects, implying a high frequency of multiple cancers. Eleven (69%) out of the 16 tumours were of squamous cell type (including carcinoma in situ), this being twice as frequent as in age-matched controls. A further characteristic was predominance in the central part of lung (69%). The lung Cr burden was very much higher [40-15,800 micrograms g-1 (dry)] in patients with lung tumours than in those without (8-28 micrograms g-1). Five of the precancerous lesions followed by bronchoscopy originated at bronchial bifurcations. Four of these cases showed a return to normal histology at autopsy even without therapy, and the other did not progress.

Most of the previous studies were epidemiological, and there are a few papers which have studied the pathology. The available information on the effects of Cr concentration in the lung on carcinogenesis is limited (Baetjer et al., 1959;Hueper, 1966;Tsuneta et al., 1980). Though chromate workers provide a good model for bronchial carcinogenesis, no detailed observations have been reported of precancerous lesions in the bronchial epithelium. We studied 13 autopsy cases to address these problems.

Subjects
Since October 1975 we have followed up a population of 84 males employed in a Tokyo factory which produced chromate compounds until August 1975. All of the 84 exworkers had chest radiograph and sputum cytology examinations in a local hospital every 6-12 months. On the detection of abnormal radiological shadows, or atypical cells on sputum cytology, patients were admitted to our hospital. Further tests such as bronchoscopy with biopsy and computerised tomography (CT) (available since 1980) were performed. In subjects with cancer, appropriate therapies including surgical intervention, irradiation and chemotherapy were applied. Wherever possible, autopsy examinations were performed. Patient 8 did not present at hospital until the onset of cancer, while the other 12 patients belonged to the group of 84 ex-workers. Epidemiological studies in these 84 ex-workers have been published elsewhere (Nakagawa et al., 1984). The present report concerns 13 male autopsies carried out consecutively before June 1993 (Table I).
Collection of data on Cr exposure and smoking Histories of exposure to chromate compounds and cigarette smoke were obtained by clinicians (K.N. and T.H.) directly from the subjects while they were treated in their local or our hospital. For other ex-chromate workers who were treated elsewhere, a volunteer worker who kept in touch with their families as well as the subjects provided us with the relevant exposure data. When calculating durations of exposure to Cr, periods of military service or non-exposed jobs such as office work were excluded. Thus, the duration only included periods of actually working in a dusty environment. Cumulative smoking dose was expressed as 'pack-years', a product of numbers of packs (20 cigarettes) a day and the duration of smoking in years. Chromium determination Samples of 3-5 g of dry tissue were taken from normal peripheral regions of lungs. The samples were wet ashed with a mixture of nitric, sulphuric and perchloric acids. The Cr content was measured by Professor K. Takemoto (Saitama Medical College, Saitama, Japan) using a flameless atomic absorption spectrophotometer (Varian Techtron, Australia, CRA 63 and GTA 95 fitted to AA 1000, 1100 and 1150). Control samples of normal lung tissues were obtained from autopsies of three non-chromate workers with lung cancer at the Cancer Institute. These patients were identified as suitable sex-and age-matched controls for the ex-chromate workers for whom Cr concentration was measured (Table II).
The smoking histories of the controls were obtained in the same manner as for the chromate cases.
Histological examination When abnormal findings such as redness or protuberance of bronchial mucosa were detected by bronchoscopy during follow-up, biopsies were performed and the lesions were classified histologically by two pathologists (Y.I. and E.T.) according to a WHO system (WHO, 1981). Dysplasia was divided into three grades (slight, moderate and severe) by the degree of cellular and structural atypia. This grading is similar to that used in cervical dysplasia (WHO, 1975). At autopsy, bronchial tissue was collected and prepared for histological examination. Specimens of 3-5 mm were cut tal bronchi), consequently a large number of blocks were made. The sections were stained with haematoxylin and eosin and, as necessary, with special stains such as periodic acid-Schiff (PAS), alcian blue, elastica van Gieson and Masson's trichrome.
Primary tumours were differentiated from metastases in cases of multiple cancer. The cell type, time of occurrence, degree of advancement, localisation of main tumour, responsiveness to irradiation and results of special staining were recorded.
For comparison of the histological types of lung cancer, a sex-and age-matched control group (n = 24,635), compiled by Morita and Sugano (1990) from pathological autopsy cases from the Japanese general population performed during 1978-87, was referred to. For the purpose of comparison with controls or foreign populations, carcinoma in situ (CIS) was included in the squamous cell carcinoma (SCC) category. Since the smoking habits of the control group were not available from the paper, the general smoking status of Japanese males was taken from the literature. For statistical analysis, the chi-square test was employed.

Results
Exposure to Cr compounds and cigarette smoke The histories of exposure to chromate compounds for the 13 autopsy cases are given in Table I Table I. Of the 13 subjects, three (nos. 3, 4 and 11) were non-smokers and two (nos. 8 and 13) stopped smoking more than 20 years before the discovery of their carcinomas. This gave a smoking prevalence of 62% at the start of follow-up and a 77% proportion of persons with smoking experience. The smoking prevalence in Japanese adult males was 84% at highest in 1966 (Mizuno & Akiba, 1989). Cumulative exposure was 28.1 pack-years on average (n = 9). Data for the control cases for Cr determination are given in Table  II. Cr concentration in the lung Lung Cr levels are shown in Table I for chromate and Table  II for controls. The values for ex-chromate workers ranged from 8 to 15,800 [jLg g-(dry)] (n = 8). Control values were 3.7 to 10.0 (average 6.1 g g-'). For subject no. 10, whose Cr level was extraordinarily high, specimens were taken from three different parts of the lungs and the measurements were repeated, giving similar results. Additionally, the lung tissues of this subject also showed high concentrations of Fe, Ni, Mg, Ni, Co, and Mn, which were uncommon in the other ex-chromate workers.
In relation to smoking, the lowest chromate level was found in a moderate smoker (28 pack-years), and nonsmokers had lower values. In the three controls, the lowest value was in a heavy smoker (50 pack-years). These data suggest that the Cr levels do not depend upon smoking habits. Table III, during follow-up and at autopsy, a total of 21 tumours were observed in 12 subjects. They comprised 16 lung, one maxillary, one oesophageal, one common bile duct and two gastric cancers. Of the 21 carcinomas, 17 arose in the respiratory tract. One subject (no. 3) with low Cr level had no cancer and died of brain infarction.

Differentiation of metachronous lung twnours
Six cases of metachronous multiple lung tumours were found (nos. 1, 2, 5, 6, 9 and 10) (Tables III and IV). The tumours were either of a different histological type (nos. 2 and 5), or showed only early submucosal invasion, making metastatic spread unlikely (nos. 1, 5 and 6). In one subject with a 3 cm tumour (no. 9), no lymph node metastasis was seen. The (n = 10) (n = 9) (n = 9) NE, not evaluated.  second cancer was on the opposite side and its spread was limited. At autopsy, no residual tumour was seen at either resection site. In subject 10, the first tumour was very small and suggested to be an early cancer by CT. After irradiation, no residual cells were seen at autopsy (see Tables III and  IV. For pulmonary tumour nodules other than those listed in Table III, the possibility that they might be metastatic cannot be ruled out completely and, hence, they were not included in the multiple cancer category. Therefore, the number of multiple cancers is a conservative estimate. The number of blocks used for histological examination ranged from 45 to 208, depending on the tissue availability (Table III).
Histological types of chronate lung cancers In the 13 autopsies, nine subjects showed 16 lung tumours. Eleven (69%) were SCC or CIS and three (19%) were SCLC, of intermediate cell type (Table III). In terms of the SCC differentiation degree, eight were moderately and one was well differentiated. The single well-differentiated SCC was in subject 12, whose Cr level was the lowest associated with lung tumour development and whose number of pack-years was the highest. In Figure 1 the histological types of lung cancer in the ex-chromate workers and the control group are compared. The prevalence of SCC (including CIS) is significantly increased in the former (P <0.005). Few adenocarcinomas were seen. In our series, the percentage of subjects with SCLC was not increased. One subject (no. 1) with a very high Cr concentration developed two SCLCs (see Figure 2).
All the subjects with lung cancer showed high Cr levels in lung (40-15,800 tg g-), whereas those without showed relatively low levels (8-28 Lg g-'). This implies a dose-response relationship between pulmonary Cr concentration and induction of lung tumours. Additionally, multiple carcinomas tended to arise in cases with higher Cr levels of lung. The four patients with the highest values showed multiple tumours.
Distribution ofprinary cancer sites in the lung The 16 lung tumours were classified as central or peripheral in origin. Eleven (69%) were central. These included an adenocarcinoma (AC) (subject 8) and a large-cell carcinoma (LCC) (subject 11).
Temporal changes in precancerous lung lesions observed during follow-up Detailed studies were made of the development of five precancerous lesions of the bronchial epithelium by bronchoscopy with biopsy. As shown in Table V, the degree of atypia varied markedly from biopsy to biopsy. For lesion 2 of subject 6, the tissue at the first biopsy was taken from a bifurcation of the right lower lobar bronchus to the medial segmental branch. Macroscopically, it showed redness and was severely dysplastic. The second bronchoscopy, 1 month after the first, showed no redness and histology demonstrated only squamous metaplasia. However, at 20 and 31 months after the first examination, severe dysplasia was again observed. Autopsy revealed only squamous metaplasia. This was similar to lesion 2 of subject 10. These findings imply that dysplasia can revert to normal or squamous metaplasia without any therapy. Furthermore, lesion 3 of subject 6, diagnosed as SCC with macroscopic redness, disappeared after irradiation, no atypical cells being observed except at the biopsy performed immediately after the irradiation. One of the most remarkable characteristics is that all five precancerous lesions occurred at bronchial bifurcations.
The present autopsy study confirmed the earlier epidemiological findings of a strong causative link between Cr exposure and lung tumours. The pathological characteristics of the tumours are as follows: 1. an increased incidence of multiple cancers composed of SCC (including CIS) and/or SCLC; (2) (1) 2. a predominance of SCC; 3. a tendency for an origin in the central part of lung. The lung tissue Cr concentrations were also much higher in . the chromate workers with lung tumours than in those with--0 out or the controls. Atypical pulmonary lesions were mainly seen at bronchial bifurcations, where Cr is known to be 0 preferentially deposited (Ishikawa et al., 1994). However, our findings indicate that such precancerous lesions may revert to normal or squamous metaplasia without any therapeutic procedure. 0

Sysstemic tumnours
No evidence was gained for any link between the common bile duct carcinoma in subject 4 and exposure to Cr. The lung Cr content in this case was one of the lowest found. On the other hand, the oesophageal carcinoma of subject 5 could°b e related to Cr inhalation because this subject was heavily exposed and inhaled Cr can accumulate in oesophageal tis-0 sue. Interestingly, subjects I I and 13 developed stomach cancers which may be sequelae of Cr exposure. However, the link remains uncertain because the stomach is the most common site of cancer occurrence in Japan and because one of the subjects seemed to have been only slightly exposed to Cr 0 (no. 13). .
Based on 123 cases, Hueper (1966) reported the histo-M logical types of lung tumours in exposed individuals as fol-E lows: SCLC, 37%; round, anaplastic and undifferentiated cell carcinomas, 54%; and AC, 9%; as compared with 29%, 61% > and 10%, respectively in controls. He concluded 'a degree of 2 variation which roughly corresponded with that seen in 'spontaneous' cancers of this organ. This observation offers a good illustration of the fact that there is no definite relationship between any particular carcinogenic agent and any specific histologic type of cancer.' We presume the main reason why he did not detect any difference between his two groups is that only a small proportion of ACs were exhibited in both, suggesting a common determinant, presumably inhaled carcinogens. The high percentage of ACs in our controls (35%) may have allowed us to obtain significant results. lasty, we must note that the criteria used for histological typing have differed with time.
Preferential site of chromate hlg cancer Most lung tumours (69%) originate centrally in line with the report of Murao et al. (1981) that 96% (23/24) are so situated. This is due in part to the increase in the proportion of SCCs, which often arises centrally (Spencer, 1985). In this respect, studies of lung cancer undertaken in Japan may be more sensitive to the detection of any increase in the central location because the background rate is lower than in European countries and the United States. We therefore concluded that predominance of central tumours is a characteristic of chromate lung cancer. In this context, it is interesting that one subject with AC (no. 8) and one with LCC (no. 11) were observed in the central region, while, often, non-chromate AC and LCC arise at the periphery (Shimosato et al., 1982;Spencer, 1985).
The dose-response relationship found in the present study was not apparent in the series of Baetjer et al. (1959). In their study, the period between the end of exposure and the time when tissue was obtained was only 0-2 years for the five non-cancer cases, while in our 13 subjects it was around 20 years. Tsuneta et al. (1980) reported a significant relationship between Cr levels and the duration of exposure, which we also found in spite of wide scatter (data not shown); however, as evidenced by the present ngs, duration of exposure might not be a reliable quantitative index for degree of Cr exposure. The kind of manufacturing process rather than its duration might also be inportant as well as the actual environmental levels.
'Safe' levels ofpulmonary Cr concentration As there appeared to be a threshold between the Cr concentrations of the subjects with lung tumours (40-15,800 pg g ') and of those without (8-28pgg-1) which was close to the average control value of 6.1lpgg-', it might be prsumed that there exists a 'safe' level of Cr inhalation. However, the present Study is based on only 13 autopses, and many more cases would be required to confirm any concusions rding 'safe' lees. Furthermore, asessment of cancer arising in organs other than the lung is also noess , because Hueper (1966) indicated that Cr concentrations may be generally increased.

Smoking habit
The smoking prevalence of the chromate workers, 62% (or 77% when ex-smokers were included), and the mean dose per day are essentially similar to those of the Japanese general mal population.
For comparison of the effects of smoking on excess risk of lung tumours between two ethnic male populations, i.e. British and Japanese, Mizuno and Akiba (1989) proposed a statistical model for calulation of the lung cancer mortality, referring to the formula for the annual lung cancer incidence in British smokers (Doll & Peto, 1978). When applying the proposed formula to the smoling status of our subjects, assuming an identity between incidence and mortality, the mortality figures per 100,000 derived from British (Doll & Peto, 1978) and Japanese formulae were found to be 393 and 228 respectively. The 70% higher mortality in British smoking males may reflect to some extent the three times higher incidence of general male lung cancer in Britain than in Japan (Hammar, 1994). On the other hand the lower Japanese mortality implies that the smoking effect on lung cancer development is relatively small. Table I, all the SCCs and SCLCs wre observed in smokers, but it must be remembered that most of smokers had heavy exposure to Cr, as indicated by their lung Cr levels. Thus, we could not rule out the possibility that the observed SCCs and SCLCs might have been caused by combined effects of exposure to both Cr and cigarette smoke rather than exposure to Cr alone. The two tumour types are related to cigarette smoking (Kreyberg, 1969). This appears to be true particulay for subject 12, who was the kast exposed to Cr of the lung tumour patients and was the heaviest smokcer (Table I). As discussed below, however, we consider that cigarette smoke was not such a potent carcinogen as inhaled Cr. Nakagawa et al. (1984) argued that the relative risk of lung cancer morbidity in our ex-chromate worker population whose durations of exposure were 9 years or longer was 21.6.

As shown in
They also caculated the relative risk of lung cancer mortality in the general male population, with the same smoking habit as our chromate workers, to be 3-4.7. Although there may be a slight difference between morbidity and mortality, these figures thus show Cr inhalation is a much more potent carcinogen than smoking. However, we could not use mortality in our case, because long-term, more intensive care has been taken of the workers than of the general population to detect tumours.
When updated epidemiologial results become available in due course, the question of whether the effects of Cr inhalation and of smoking are additive or multiplicative will be analysed.

Precancerous lung lesions
The fact that the bronchi of ex-chromate workers showed an increased Prevakln of SCC and CIS or dysplasia is consided to be one of their characteristic features. In contast, the pevakln of CIS associated with primary lung cancer cases in the non-chromate Japanese male population is very low, only 1/59 or 1.7% (Wakimoto et al., 1982), although that of squamous metaplasia is 52/81 or 64% (by cell type, 83% in SCC, 84% in SCLC and 43% in AC patients) (Tsuchiya et al., 1987). As discussed in these reports, the Prvaklc of CIS in the Japanese population are much lower than in the United States (82.5% according to Auerbach et al., 1961;13% according to Valentine, 1957), probably rLecting their lower ovenll percentage of SCC as primary hmg tumours of Japanese population.
The possibility that precancerous lesions might revert spontaneously to normal ciliated epithelium or to squamous metaplasia was sug d by the present data. However, alternative interpretations are as follows: 1. The biopsy which led to the detection of dysplasias might have removed all of the altered tissue. 2. The precancerous lesions might have consisted of multiple cell groups with different degrees of atypia, if any, and each biopsy might pickc up only a single population. 3. Associated inflammation might augment macroscopic atypia (i.e. redness) and histological grading (prticularly in the case of slight dysplasia as a result).
The first is possible because multiple specimens were sometimes sampled even from a small lesion. On the other hand, the second explanation means that the possibility of residual unidentified atypical cells requires consideration. It should be noted that macroscopic redn is often seen in association with atypical cells and may indeed be an inflammatory reaction to their presence. If this is the case, the lack of the redness in association with negative biopsy specimens would imply actual disappearance. However, the third possibility must also be taken into account in further, more comprehensive, studies, using a greater number of lesions, to improve our understanding. We thank Professor K. Takemoto, Saitama Medical College, for measurement of Cr in the lung tissues and Dr A. Morgan, AEA Biomedical Research, for useful comments and assistance in the text preparation. We are also indebted to Professor H. Shimizu, Gifu University, for his useful advice on epidemiology. A part of this study was supported financially by the Smoking Research Foundation, Japan, and the AEA Biomedical Research, UK.