Vol. 19, No.3 (5uppl. 1) Printed in Great Britain
International Journal of Epidemiology
© International Epidemiological Association 1990
Histopathology of Lung Cancer DAVID T CARR
given a tiny biopsy specimen that may not be truly representative of the entire tumour. In such cases, a poorly differentiated epidermoid carcinoma or adenocarcinoma may be misclassified as a large cell, undifferentiated carcinoma.
CELL TYPES Lung cancer is at present an important disease worldwide-in the US the number of new cases in 1989 was 155 000 with 142000 deaths.' Incidence rates vary between populations and between males and females, suggesting that environmental factors play an important role in the establishment of the disease. From a pathologist's point of view, lung cancer can be classified according to different histological cell types. Table 1 reports a commonly used histological classification of malignancies of the lung. Of the different types in the table, epidermoid and adenocarcinomas account for the largest proportion and approximately 90% of all lung carcinomas are included in the four main histopathological categories: squamous cell, small cell, adenocarcinoma and large cell carcinoma. From an epidemiological point of view, it has been proposed that the tumours be classified into two groups where group 1 tumours (squamous cell and small cell carcinomas) are usually located centrally in the respiratory tract and group 2 tumours peripheral in location and comprising bronchoalveolar carcinomas, adenocarcinomas and large cell carcinomas.' Establishing the correct diagnosis in a single case is sometimes very complicated. Cases of epidermoid carcinoma and adenocarcinoma may be well differentiated, easily recognized cases, or they may be poorly differentiated and the pathologist may have difficulty finding any evidence that would enable him to classify properly the tumour. Furthermore, he may have been
ORIGIN OF LUNG CANCER The lung is formed from an endodermal bud which grows into a tube lined by a single layer of endodermal cells. These give rise to all of the cells lining the respiratory tree--ciliated as well as secretory cells. From a morphological point of view, it has been suggested that all forms of lung cancer, including carcinoids and small cell carcinomas are derived from this primitive endoderm." This concept was reviewed by Yesner' in 1981. Based upon histopathological examinations, including special stains and routine electron microscopy of lung tumours, he suggested that all the tumours derived from the bronchial epithelium are related. Differences in histological appearance represent a dynamic spectrum with development towards or away from small cells, squamous cells, or adenocarcinomas. The conditions under which the different histological types appear are dependent on the type of environmental exposure, host responses and, in clinical cases on therapy. TABLE
Male Female Total
Department of Internal Medicine, MD Anderson Hospital and Tumor Institute, University of Texas System Cancer Center, Houston, Texas, USA.
1
Lung cancer-1989*
New cases
Deaths
101000 54000 155000
93000 49000 142000
* Estimates of American Cancer Society'
S8
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
Carr DT (Department of Internal Medicine, MD Anderson Hospital and Tumor Institute, University of Texas System Cancer Center, Houston, Texas, USA). Histopathology of lung cancer. International Journal of Epidemiology 1990,19 (SuppI1): 58-510. Lung cancer is a complex problem because there are a number of different histological cell types. Those commonly grouped as bronchogenic carcinoma (epidermoid carcinoma, adenocarcinoma, large cell undifferentiated carcinoma, small cell carcinoma, and adenosquamos carcinoma) account for more than 90% ofthe new cases and the deaths each year. The natural history of bronchogenic carcinoma suggests that many years pass while the cancer evolves from a precancerous change in the bronchial mucosa, to undetectable microscopic cancer, to preclinical asymptomatic cancer and finally into a full symptomatic cancer, the phase of most lung malignancies in the tissue at diagnosis. Therefore, students of the aetiology ofthis disease must consider what has happened to patients 5-20 years before lung cancer is diagnosed.
HISTOPATHOLOGY OF LUNG CANCER
Diagnosis The skill and experience of the pathologist is obviously a factor in the accuracy of the histological classification of cases, especially in difficult cases. There have been reports of significant inter-observer variability and even intra-observer variability in classification of cases by histological cell type, when the same slides were examined by different experts and later re-examined as unknowns by the same experts. In one study," five experienced pathologists gave two independent readings of 50 different specimens. Disagreement by individual pathologist from the consensus reading, varied from 2% to 42% depending on the degree of differentiation of the cancers. Differences in the separate readings of the same slides by the same pathologist varied from 2% to 20%. In addition, many cancers have two or more histological pictures in various parts of the tumour, and it
must be remembered that the classification is based on the predominant histological type. 9,10 This combination of two or more histological types in the same tumour probably accounts for a part of the observed changes during the various phases of treatment of some patients. For example, a patient with small cell cancer may respond to chemotherapy but have roentgenographic evidence of residual disease at the site of the primary tumour which on resection will be found to be either squamous cell carcinoma, or adenocarcinoma without any residual small cell cancer. In such cases it may be presumed that the patient initially had a mixture of cell types and that the chemotherapy eradicated the small cell component, leaving the non-small cell portions which were not as sensitive to the chemotherapy. I! Accurate histological classification of all cases may be of special importance to the epidemiologist as there may be important correlations between aetiology and the cell type of the tumour." For example, chloromethyl ether is known to cause small cell carcinomas'i' and it has been reported that small cell carcinoma is a more common type in uranium miners than in the general population. 14 SOURCE OF DATA Incidence rates for lung cancer and death rates attributed to that disease are based on many different sources of information varying from physicians' reports to hospital cancer registers to state cancer registers and mortality statistics based on death certificates. All these data have a substantial built in rate of error due to the difficulties in differentiating cases of lung cancer from cases of other pulmonary disease and differentiating primary lung cancer from cancer that has metastasized from other primary sites to the lungs. Some of the data do not include a classification by cell type of the cancer. There has been a marked decrease in the rate of autopsies in recent decades and this has undoubtedly resulted in some decrease in accuracy of statistics on both incidence and death rates attributable to lung cancer. A recent report" from Yale University School of Medicine indicated that the correct diagnosis of lung cancer had not been made in 26% of 153 patients who had an autopsy between 1971 and 1982. In Sweden, about 15% of the lung cancer cases are incidentally found at autopsy, again demonstrating the potential influence autopsy rates on lung cancer incidence figures. NATURAL HISTORY AND LATENCY TIME Studies of the causes of lung cancer must take into con-
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
Epidemiological evidence suggests that cigarette smoking increases the incidence of all four major histological cell types," whereas for certain occupational exposures, distinct cell types are increased. Small cell carcinomas seem to dominate bis (chloromethyl) ether exposure," although the histopathological specificity in these cases may also be determined by factors not related to the exposure. Important evidence regarding the concept of transition between different histological types has been reviewed by McDowell and Trump" who summarized their own work on transitional changes in animals devoid of vitamin A. Apart from vitamin A, the concentration of calcium in in vitro preparations has been found to influence proliferation, stratification and keratinization and an apparent genetic instability in tumour cells has been demonstrated following transfer of tumour cells in new hosts. The importance of hormones, illustrated by the difference in incidence of adenocarcinomas between males and females, has been verified in many studies. From an epidemiological point of view, the study of shifts in relative frequency of different tumour types may provide an interesting epidemiological tool. Such studies have been published e.g. on lung cancer in New Mexico." In that particular study, a decrease in the proportion of adenocarcinomas was found among females between the periods of 1970-72 and 1980-81. The study demonstrated a change in diagnostic criteria applied by the pathologists between the two time periods. This adds to the uncertainties of performing epidemiological studies using histopathology in addition to the observer errors and normal variation in histological types as discussed above.
S9
SIO
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY (SUPPLEMENT 1)
CONCLUSIONS Although seemingly accurate, the histopathological spectrum of lung cancer provides many pitfalls for the
epidemiologist interested in the influence of environmental agents on specific cell types. A high autopsy rate, good precision in the histopathological examination with several specimens taken from the same tumour and careful control of observer variables are prerequisites for successful studies. Unless these requirements are met, regional, temporal or exposure group related differences in histopathological types should be evaluated with caution. REFERENCES Silverberg E, Lubera JA. Ca-A Cancer Journal for Clinicians 1989; 39: 3-20. 2 Kreyberg L. Histological lung cancer types. A morphological and biological correlation. Acta Pathol Microbial Scand 1962; 157: 1-93. 'Ycsner R. The dynamic histopathologic spectrum of lung cancer. Yale J BioI Med 1981; 54: 447-56. 4 U.S. Department of Health and Human Service. The health consequences of smoking: Cancer. A report of the Surgeon General. Washington. D.C. U.S. Govt Print Off 1982 (DHHS Publication No. 82-5(279). 5 Ives JC. Buffler PA, Greenberg SO. Environmental associations and histopathologic patterns of carcinoma of the lung: The challenge and dilemma in epidemiologic studies. Am Rev Respir Dis 1983; 128: 195-209. (, McDowell EM, Trump BF. Histogenesis of preneoplastic lesions in tracheobronchial epithelium. Survey Synthesis Pathol Res 1983; 2: 235-79. 7 Butler C. Samet JM. Humble CG. Histopathology of lung cancer in New Mexico. 1970-72 and 1980-81. JNCl1987; 78: 85-90. x McFarlane MJ. Feinstein AR. Wells CK. Necrospy evidence of detection bias in the diagnosis of lung cancer. Arch Intern Med 1986; 146: 1695-98. 9 Weiss W. Implications of tumor growth rate for the natural history of lung cancer. J Occup Med 1984; 26: 345-52. \lJ Carter D. Precancer in the lung. Cancer Surveys 1983; 2: 425-34. 1
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
sideration the natural history of the disease." Lung cancer is a chronic disease with the passage of many years as it evolves through the many phases of its development from premalignant changes in the bronchial mucosa to the various phases of the cancer. Most lung cancers are not diagnosed until they reach the clinical or symptomatic phase by routine roentgenograms of the thorax, the lower limit of deductibility being a mass about one centimetre in diameter. Such a mass contains about one billion cancer cells. We may assume that such cancer begins as a single cancer cell and doubles and redoubles until the mass of cells reaches one billion cells. Thirty doublings are required to go from one to one billion. The usual doubling time for a mass of lung cancer cells varies from about one month for the small cell variety to three to six months for the non-small cell types. If we use three months as a common doubling time for a mass of lung cancer cells, we can calculate that 7.5 years may pass as the cancer grows and reaches the detectable phase in its development. Another year or longer may pass before it becomes symptomatic which is the usual phase in which most lung cancers are diagnosed. Furthermore, the premalignant period has been estimated to vary from one to ten years. III Therefore, the aetiologist must be particularly concerned about environmental and host factors present 5-20 years prior to the date of diagnosis of the lung cancer.
International Journal of Epidemiology
© International Epidemiological Association 1990
Histopathology of Lung Cancer DAVID T CARR
given a tiny biopsy specimen that may not be truly representative of the entire tumour. In such cases, a poorly differentiated epidermoid carcinoma or adenocarcinoma may be misclassified as a large cell, undifferentiated carcinoma.
CELL TYPES Lung cancer is at present an important disease worldwide-in the US the number of new cases in 1989 was 155 000 with 142000 deaths.' Incidence rates vary between populations and between males and females, suggesting that environmental factors play an important role in the establishment of the disease. From a pathologist's point of view, lung cancer can be classified according to different histological cell types. Table 1 reports a commonly used histological classification of malignancies of the lung. Of the different types in the table, epidermoid and adenocarcinomas account for the largest proportion and approximately 90% of all lung carcinomas are included in the four main histopathological categories: squamous cell, small cell, adenocarcinoma and large cell carcinoma. From an epidemiological point of view, it has been proposed that the tumours be classified into two groups where group 1 tumours (squamous cell and small cell carcinomas) are usually located centrally in the respiratory tract and group 2 tumours peripheral in location and comprising bronchoalveolar carcinomas, adenocarcinomas and large cell carcinomas.' Establishing the correct diagnosis in a single case is sometimes very complicated. Cases of epidermoid carcinoma and adenocarcinoma may be well differentiated, easily recognized cases, or they may be poorly differentiated and the pathologist may have difficulty finding any evidence that would enable him to classify properly the tumour. Furthermore, he may have been
ORIGIN OF LUNG CANCER The lung is formed from an endodermal bud which grows into a tube lined by a single layer of endodermal cells. These give rise to all of the cells lining the respiratory tree--ciliated as well as secretory cells. From a morphological point of view, it has been suggested that all forms of lung cancer, including carcinoids and small cell carcinomas are derived from this primitive endoderm." This concept was reviewed by Yesner' in 1981. Based upon histopathological examinations, including special stains and routine electron microscopy of lung tumours, he suggested that all the tumours derived from the bronchial epithelium are related. Differences in histological appearance represent a dynamic spectrum with development towards or away from small cells, squamous cells, or adenocarcinomas. The conditions under which the different histological types appear are dependent on the type of environmental exposure, host responses and, in clinical cases on therapy. TABLE
Male Female Total
Department of Internal Medicine, MD Anderson Hospital and Tumor Institute, University of Texas System Cancer Center, Houston, Texas, USA.
1
Lung cancer-1989*
New cases
Deaths
101000 54000 155000
93000 49000 142000
* Estimates of American Cancer Society'
S8
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
Carr DT (Department of Internal Medicine, MD Anderson Hospital and Tumor Institute, University of Texas System Cancer Center, Houston, Texas, USA). Histopathology of lung cancer. International Journal of Epidemiology 1990,19 (SuppI1): 58-510. Lung cancer is a complex problem because there are a number of different histological cell types. Those commonly grouped as bronchogenic carcinoma (epidermoid carcinoma, adenocarcinoma, large cell undifferentiated carcinoma, small cell carcinoma, and adenosquamos carcinoma) account for more than 90% ofthe new cases and the deaths each year. The natural history of bronchogenic carcinoma suggests that many years pass while the cancer evolves from a precancerous change in the bronchial mucosa, to undetectable microscopic cancer, to preclinical asymptomatic cancer and finally into a full symptomatic cancer, the phase of most lung malignancies in the tissue at diagnosis. Therefore, students of the aetiology ofthis disease must consider what has happened to patients 5-20 years before lung cancer is diagnosed.
HISTOPATHOLOGY OF LUNG CANCER
Diagnosis The skill and experience of the pathologist is obviously a factor in the accuracy of the histological classification of cases, especially in difficult cases. There have been reports of significant inter-observer variability and even intra-observer variability in classification of cases by histological cell type, when the same slides were examined by different experts and later re-examined as unknowns by the same experts. In one study," five experienced pathologists gave two independent readings of 50 different specimens. Disagreement by individual pathologist from the consensus reading, varied from 2% to 42% depending on the degree of differentiation of the cancers. Differences in the separate readings of the same slides by the same pathologist varied from 2% to 20%. In addition, many cancers have two or more histological pictures in various parts of the tumour, and it
must be remembered that the classification is based on the predominant histological type. 9,10 This combination of two or more histological types in the same tumour probably accounts for a part of the observed changes during the various phases of treatment of some patients. For example, a patient with small cell cancer may respond to chemotherapy but have roentgenographic evidence of residual disease at the site of the primary tumour which on resection will be found to be either squamous cell carcinoma, or adenocarcinoma without any residual small cell cancer. In such cases it may be presumed that the patient initially had a mixture of cell types and that the chemotherapy eradicated the small cell component, leaving the non-small cell portions which were not as sensitive to the chemotherapy. I! Accurate histological classification of all cases may be of special importance to the epidemiologist as there may be important correlations between aetiology and the cell type of the tumour." For example, chloromethyl ether is known to cause small cell carcinomas'i' and it has been reported that small cell carcinoma is a more common type in uranium miners than in the general population. 14 SOURCE OF DATA Incidence rates for lung cancer and death rates attributed to that disease are based on many different sources of information varying from physicians' reports to hospital cancer registers to state cancer registers and mortality statistics based on death certificates. All these data have a substantial built in rate of error due to the difficulties in differentiating cases of lung cancer from cases of other pulmonary disease and differentiating primary lung cancer from cancer that has metastasized from other primary sites to the lungs. Some of the data do not include a classification by cell type of the cancer. There has been a marked decrease in the rate of autopsies in recent decades and this has undoubtedly resulted in some decrease in accuracy of statistics on both incidence and death rates attributable to lung cancer. A recent report" from Yale University School of Medicine indicated that the correct diagnosis of lung cancer had not been made in 26% of 153 patients who had an autopsy between 1971 and 1982. In Sweden, about 15% of the lung cancer cases are incidentally found at autopsy, again demonstrating the potential influence autopsy rates on lung cancer incidence figures. NATURAL HISTORY AND LATENCY TIME Studies of the causes of lung cancer must take into con-
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
Epidemiological evidence suggests that cigarette smoking increases the incidence of all four major histological cell types," whereas for certain occupational exposures, distinct cell types are increased. Small cell carcinomas seem to dominate bis (chloromethyl) ether exposure," although the histopathological specificity in these cases may also be determined by factors not related to the exposure. Important evidence regarding the concept of transition between different histological types has been reviewed by McDowell and Trump" who summarized their own work on transitional changes in animals devoid of vitamin A. Apart from vitamin A, the concentration of calcium in in vitro preparations has been found to influence proliferation, stratification and keratinization and an apparent genetic instability in tumour cells has been demonstrated following transfer of tumour cells in new hosts. The importance of hormones, illustrated by the difference in incidence of adenocarcinomas between males and females, has been verified in many studies. From an epidemiological point of view, the study of shifts in relative frequency of different tumour types may provide an interesting epidemiological tool. Such studies have been published e.g. on lung cancer in New Mexico." In that particular study, a decrease in the proportion of adenocarcinomas was found among females between the periods of 1970-72 and 1980-81. The study demonstrated a change in diagnostic criteria applied by the pathologists between the two time periods. This adds to the uncertainties of performing epidemiological studies using histopathology in addition to the observer errors and normal variation in histological types as discussed above.
S9
SIO
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY (SUPPLEMENT 1)
CONCLUSIONS Although seemingly accurate, the histopathological spectrum of lung cancer provides many pitfalls for the
epidemiologist interested in the influence of environmental agents on specific cell types. A high autopsy rate, good precision in the histopathological examination with several specimens taken from the same tumour and careful control of observer variables are prerequisites for successful studies. Unless these requirements are met, regional, temporal or exposure group related differences in histopathological types should be evaluated with caution. REFERENCES Silverberg E, Lubera JA. Ca-A Cancer Journal for Clinicians 1989; 39: 3-20. 2 Kreyberg L. Histological lung cancer types. A morphological and biological correlation. Acta Pathol Microbial Scand 1962; 157: 1-93. 'Ycsner R. The dynamic histopathologic spectrum of lung cancer. Yale J BioI Med 1981; 54: 447-56. 4 U.S. Department of Health and Human Service. The health consequences of smoking: Cancer. A report of the Surgeon General. Washington. D.C. U.S. Govt Print Off 1982 (DHHS Publication No. 82-5(279). 5 Ives JC. Buffler PA, Greenberg SO. Environmental associations and histopathologic patterns of carcinoma of the lung: The challenge and dilemma in epidemiologic studies. Am Rev Respir Dis 1983; 128: 195-209. (, McDowell EM, Trump BF. Histogenesis of preneoplastic lesions in tracheobronchial epithelium. Survey Synthesis Pathol Res 1983; 2: 235-79. 7 Butler C. Samet JM. Humble CG. Histopathology of lung cancer in New Mexico. 1970-72 and 1980-81. JNCl1987; 78: 85-90. x McFarlane MJ. Feinstein AR. Wells CK. Necrospy evidence of detection bias in the diagnosis of lung cancer. Arch Intern Med 1986; 146: 1695-98. 9 Weiss W. Implications of tumor growth rate for the natural history of lung cancer. J Occup Med 1984; 26: 345-52. \lJ Carter D. Precancer in the lung. Cancer Surveys 1983; 2: 425-34. 1
Downloaded from http://ije.oxfordjournals.org/ at East Carolina University on June 27, 2015
sideration the natural history of the disease." Lung cancer is a chronic disease with the passage of many years as it evolves through the many phases of its development from premalignant changes in the bronchial mucosa to the various phases of the cancer. Most lung cancers are not diagnosed until they reach the clinical or symptomatic phase by routine roentgenograms of the thorax, the lower limit of deductibility being a mass about one centimetre in diameter. Such a mass contains about one billion cancer cells. We may assume that such cancer begins as a single cancer cell and doubles and redoubles until the mass of cells reaches one billion cells. Thirty doublings are required to go from one to one billion. The usual doubling time for a mass of lung cancer cells varies from about one month for the small cell variety to three to six months for the non-small cell types. If we use three months as a common doubling time for a mass of lung cancer cells, we can calculate that 7.5 years may pass as the cancer grows and reaches the detectable phase in its development. Another year or longer may pass before it becomes symptomatic which is the usual phase in which most lung cancers are diagnosed. Furthermore, the premalignant period has been estimated to vary from one to ten years. III Therefore, the aetiologist must be particularly concerned about environmental and host factors present 5-20 years prior to the date of diagnosis of the lung cancer.
