A large cell carcinoma (LCC) is one of the typical histologic types of lung cancer, but their incidence is lower than 10%, which is relatively less frequent than other types¹⁾. With the use of light microscopy they are defined as large malignant epithelial tumors, with large nuclei, prominent nucleoli and abundant cytoplasm, and they usually have well defined cell borders, without the characteristic features of squamous or small carcinomas or adenocarcinomas²⁾. However, because they are difficult to distinguish from other undifferentiated types, other then by surgical resection, their prevalence varies between 6–31% according to reporters, and consequently studies about their definite clinical characteristics and survival rates are lacking^{3, 4)}.

With the use of electron microscopy and immunohistochemical staining method, a LCC, with neuroendocrine features, was recently separated as a large cell neuroendocrine carcinoma, with a reported prognosis worse than that of a LCC itself, further augmenting the confusion and debate relating to the diagnosis of LCC^{5, 6)}.

Attempts were made to find the concordance rate between the pre- and post-operative diagnoses, clinical characteristics and survival rates of patients diagnosed with a large cell carcinoma, between May 1986 and June 2001, at the Severance Hospital, Yonsei University College of Medicine.

5,069 cases of primary lung cancer were diagnosed, between May 1986 and June 2001, at the Severance Hospital, Medical College, Yonsei University, with 1,003 being surgically resected (Table 1). Of the 31 confirmed LCC cases, 28 were analyzed, as 2 cases, re-diagnosed by pathological re-evaluation as adenocarcinoma and squamous cell carcinoma, and 1 where only an open lung biopsy was performed, were excluded.

For the pathological re-evaluation, the tissue was fixed in 10% formalin, embedded in paraffin, cut into 5-μm sections and stained with Hematoxylin-Eosin. When differentiation of the adenocarcinoma was, Periodic acid-Schiff (PAS) and mucicarmine stain were used, which was classified as a LCC if the results were negative.

Basaloid and large cell neuroendocrine carcinomas especially, were differentiated using neuron-specific enolase (1:100; DAKO, Santa Barbara, CA, USA), chromogranin A (1:10; Boehringer Mannheim, Indianapolis, IN), synaptophysin (1:10; DAKO, Santa Barbara, CA, USA) and CD56 (1:50; DAKO, Santa Barbara, CA, USA).

By reviewing the medical records, sex, age, smoking history, symptom, preoperative diagnosis, findings of initial chest computerized tomography, pre- and post-operative stages, pathological findings, postoperative treatment, the incidence of recurrence, follow-up, and survival rates were sought.

According to the 1967 WHO criteria, amended in 1982, a LCC is defined as “a malignant epithelial tumor, with large nuclei, prominent nucleoli and abundant cytoplasm, usually with well defined cell borders, but without the characteristics features of squamous or small cell carcinomas or adenocarcinomas”. Undifferentiated tumors, with intracytoplsmic mucin, which had previously been categorized as LCC, have been re-categorized as adenocarcinomas²⁾. Even though immunohistochemical staining and electron microscopy can be used to detect features of differentiation (squamous, glandular, neuroendocrine), the diagnosis of a LCC is mainly based on routine histochemical staining methods and the light microscopic findly²⁾. Even with these diagnostic criteria, it is not easy to distinguish a poorly differentiated LCC from a poorly differentiated squamous cell carcinoma or adenocarcinoma.

Recently, Travis et al⁵⁾ suggested that a non-small cell carcinoma of the lung, viewed as a neuroendocrine differentiation by light microscopy, and confirmed by immunohistochemical stain and electron microscopy, but not categorized as another neuroendocrine tumor, such as carcinoid, atypical carcinoid, and small cell carcinoma, should be named a large cell neuroendocrine carcinoma, on the basis of the reports suggesting such tumors have a poor prognosis. The WHO/IASLC lung cancer classification of 1999 defined a large cell neuroendocrine carcinoma as a variant of a LCC⁷⁾.

Our relative prevalence of a LCC was 1.7%, 85 out of 5,069 lung cancers, which was lower than the reports of other Countries, including one by Travis et al. that reported a prevalence of about 10%^{1, 3, 4)}, which was also lower than the 6.3% prevalence reported by Kim et al⁸⁾. Even if only the surgically resected primary lung cancers were considered, the 2.9% prevalence was still lower than the 3.5 and 7.6% reported by Ham et al⁹⁾ and Han et al¹⁰⁾. Our suggestions for this lower prevalence, compared to previous reports, are as followed; firstly, since 1982 some LCC were reclassified as mucinous adenocarcinoma. Secondly, if not surgically resected, a LCC could have been misdiagnosed as another type due to central necrosis and hemorrhage. A report by Kim et al.⁸⁾ analyzed the primary lung cancers that occurred between 1957 and 1977, and at that time the prevalence of an adenocarcinoma was 11.9%. Our report dealt with the lung cancers that occurred between 1986 and 2001, and while the prevalence of adenocarcinomas have increased to 27% since the mid-1980s, undifferentiated cancers, including LCC, decreased, possibly due to the wider application of immunohistochemical staining methods.

Table 2 summarizes the clinical characteristics of our patients. The gender ratio was 8.3:1, with a male predominance, which was similar to the 10.5:1 ratio reported by Han et al¹⁰⁾, but different from the 3.9:1 reported by Kim et al⁸⁾, which was based on all lung cancers. Travis et al reported a 2.2:1 ratio⁵⁾.

The average smoke index of the 22 smokers was 33 pack years. Of the 25 male patients, 3 were non-smokers, 21 (88%) had a positive history of smoking and 1 had an unknown history. Compared to the study of Barbone et al¹¹⁾, where 95% of the male patients with a LCC were reportedly smokers, ours showed a slightly lower rate. It is unknown whether smoking was associated with the predominance of LCC prevalence in male patients, and to find other involved factors involved more studies will be required.

Our study revealed that malignant cells were found in only 1 out of 25 cases using the sputum cytology method; with a diagnostic rate was 4%. According to Gledhill et al.¹²⁾ the low diagnostic rate of the sputum cytology method was due to a too smaller amount of sputum and number of diagnostic trials. To overcome this short fall, he suggested that an adequate amount of sputum should be acquired, in the morning, by coughing after a deep breath, and the tests should be repeated more than 4 times. In our study, the low diagnostic rates, even with 13 cases of central tumor and 11 cases of hemorrhage, were thought be due to the acquisition of sputum from just one test and the inaccuracy of the collection method. This emphasizes the need for repeated and accurate sputum cytology tests.

Only 10 out of 26 cases (38%) were diagnosed as lung cancer using fiberoptic bronchoscopy, with a diagnostic rate of 38%. The low diagnostic rate was because in 7 cases, where the lesion was not visible, all the biopsy samples were negative, with only 5 out of 10 cases having visible lesions, and in 5 out of 9 cases where a transbronchial biopsy was performed, was lung cancer diagnosed. Even the diagnostic rate of the visible lesion was low, as severe necrosis and hemorrhage hindered the diagnosis. Including the 2 cases where the bronchial washing cytology test was positive for malignancy, the diagnostic rate by bronchoscopy was 46.2%, i.e. 12 out of 26 cases. According to Ham et al.⁹⁾ in 189 cases where bronchoscopy was performed, the diagnostic rate was 71.4%, and according to Kamholz et al.¹³⁾ this value was 73%. Mak et al.¹⁴⁾ also reported on 125 cases where the tumor was visible with bronchoscopy, where the positive rate offor a biopsy was 76%, of the bronchial washing method was 49.6%, and of the bronchial brushing method was 52%. In 63 cases where the tumor was not visible, the corresponding figures were 36.5, 38.1, and 28.6%, respectively. From this it was recommended combining the washing and brushing methods with a biopsy during bronchoscopy. In our study, the malignancy detection rate of the bronchial washing method was 36%, 9 out of 26 cases, which was slightly lower than the reports by Mak et al¹⁴⁾. Even when the tumor was visible, a biopsy through bronchoscopy might be unable to diagnose, due to too few, or no, cancer cells in the tumor, or from tissue damage that could cause difficulty in the interpretation. Especially in LCC, central necrosis, hemorrhage and obstructive pneumonia might be anticipated in the low diagnostic rate.

The diagnostic rate of a fine needle aspiration biopsy was 88%, i.e. 14 out of 16 cases. According to Allison et al¹⁵⁾, the diagnostic rate was 90%, with almost no false positive results, and a false negative rate of 5–10%, which were compatible to our results.

Among 28 surgically resected cases, only 1 was preoperatively diagnosed as a LCC. Confusion in the diagnostic guide could be the reason for the low preoperative diagnostic rate, with central necrosis, hemorrhage and obstructive pneumonia also making diagnosis potentially difficult by methods other than surgery. For precise confirmation, an analysis of the concordance rate between preoperative diagnosis of a LCC and the postoperative confirmation of the same histologic type is required. In our literature review no such study could be found.

The tumor was sited at the periphery in 15 cases (54%), with a similar number being found centrally. Taking into account all the LCC, the peripheral type might be more abundant our study only considered the patients having undergone surgery. The tumors were located in the right lung in 18 cases (68%), which was twice as often as in the left lung. According to the report by Kim et al⁸⁾, the prevalence in the right lung was also twice that in the left lung, but reasons for this was unknown. Radiologically, our study showed that surgically confirmed LCC were distributed evenly at the periphery and center, with lobulated rather than smooth margins, with the presence of central necrosis.

Computerized tomography is most useful for diagnosing the extension of a primary tumor, its relationship to adjacent organs, mediastinal lymph node enlargement and also helps to determine the resectability. However, according to Heidenberg et al¹⁶⁾, the size of lymph node alone could not differentiate metastasis from inflammation, as not all the enlarged lymph nodes were involved with the cancer, where 87% of the involved lymph nodes were larger than 1.4 cm. With the criterion of 1.5 cm, Modini et al¹⁷⁾ achieved sensitivity, specificity and accuracy of 55, 91 and 75%, and in the report by Ham et al⁹⁾, were 78.9, 72.7 and 75%, respectively. With the criterion of 1 cm, according to Webb et al¹⁸⁾, the sensitivity and specificity were 52 and 69%, respectively, and a large scaled multivariable analysis by Dales et al.¹⁹⁾ revealed a sensitivity, specificity and accuracy each of 79%.

Our study showed that if the enlargement of the mediastinal lymph nodes were defined as larger than a diameter of 1 cm, the sensitivity was 100%, but the specificity was 52%. For the parenchymal and the hilar lymph node, the sensitivity was 20%, but the specificity and accuracy were 87 and 75%, respectively, but compared to other studies, the accuracy was still low. Pulmonary tuberculosis, which is highly prevalent in Korea, and obstructive pneumonia and central necrosis by cancer, could induce reactive lymph node hypertrophy. The concordance rate of the pre- and postoperative stage was as low as 43% due to the difference between the pathologic findings and the computerized tomography with regard to the lymph node enlargements and major organ involvements.

Generally, surgical resection is the primary treatment for a non-small cell carcinoma, including a LCC, where after radiotherapy or chemotherapy is undertaken according to the stage. Mitchell et al.²⁰⁾ analyzed the prognosis of 208 patients with a LCC, and reported that the median survival time and 5-year survival rate of 55 patients that received surgical treatment were 13 months and 21.2%, and for 4 patients that received postoperative radiotherapy were 6.2 months and 0%. Downey et al.²¹⁾ reported that even with precise surgical staging and clear diagnostic criteria, the prognosis of LCC was poor, regardless of the treatment. At the time of diagnosis, 90% of cases were already at an advanced stage (stage III, IV), with only 10% at stage I. He reported that most patients expired within 9 months’ with an average survival time for the stage I patients of 15.9 months and therefore the prognosis was very poor.

According to Downey et al²²⁾, the 5-year survival rate of 61 patients with LCC having undergone surgery was 37%, suggesting an improvementd in the results due to aggressive surgery. The 28 patients in our study had a median survival time and 5-year survival time of 54.5 months and 45%, respectively, which were higher than reports from other Countries. Compared to the report of Mitchell et al²⁰⁾, the distribution of stages were similar, but those of the postoperative chemotherapy and radiotherapy were not, and compared to the reports of Downey et al^{21, 22)}, more low stages, such as stages I and II, were included in our study (Table 9). According to a Korean report by Lee et al²³⁾, in 1987, the 3-year survival rate of 8 patients with LCC out of 147 patients with primary lung cancer that underwent a surgical resection was 44.5%, and in the report of Han et al¹⁰⁾, in 1996, the 2-year survival rate of 29 out of 382 such patients was 59%.

There have been few trials of postoperative chemotherapy and radiotherapy for LCC, suggesting multi-centered clinical studies, based on a larger population of patients, are needed.