Screening, Early Detection, and Early Intervention Strategies for Lung Cancer

Henry Wagner, Jr, MD, and John C. Ruckdeschel, MD

Thoracic Oncology Program at H. Lee Moffitt Cancer Center & Research Institute

If discoursing on a difficult problem were like carrying weights, when many horses can carry more sacks of grainthan a single horse, I would agree that many discourses would do more than a single one; but discoursing is like coursing, not like carrying, and one Barbary courser can go faster than a hundred Frieslands.

Galileo Galilei, The Tester, cited by Italo Calvino in Six Memos for the Next Millenium. NewYork, NY: Vintage Books; 1993:43.

Introduction

Lung cancer is the most frequent cause of cancer death in both men and women. While the incidence of lung cancer appears to have decreased in white men, it continues to rise in nonwhite men and in women. Most lung cancer is caused by cigarette smoking, but strategies to prevent or reduce this addiction have met with only modest success to date.[1] Smokers who quit remain at an increased though gradually declining risk of lung cancer over at least the next decade.[2] For these individuals, early detection and treatment of the cancers that they may still develop remains the best hope of further reducing their risk of death from lung cancer.

Strategies for dealing with early lung cancer fall into three themes: true screening programs for the evaluation of asymptomatic individuals believed to be at high risk for lung cancer; early detection programs for the evaluation of patients presenting with ambiguous symptoms; and early intervention programs aimed at stopping or reversing the processes involved in lung carcinogenesis before the development of invasive malignancy.

Since approximately 90% of lung cancer develops in individuals with a history of cigarette smoking, many of whom have chronic respiratory symptoms, the distinction between screening and early detection is blurred. Individuals who participate in "screening" programs may be motivated by subtle changes in their baseline cough or sputum production (or by a family member's prodding that such a change has occurred) and thus may not truly represent the larger population of smokers and ex-smokers from whom they are drawn.

Lung Cancer Presentation, Staging, and Treatment Outcome

Table 1 shows the stage distribution of 170,000 cases of lung cancer diagnosed in the United States in 1993. While a few patients with stage III disease (those with minimal weight loss and good performance status) may be cured by aggressive multimodality therapy, only those patients with stage I or II disease are classically considered to be resectable with high probability of cure. If only those patients with T1N0 disease are considered as truly early presentations, more than 85% of patients with lung cancer currently present with more advanced disease.

Table 1. -- Stage Distribution of Lung Cancer at Presentation ______________________________________________________________________

From Holmes EC. Adv Oncol. 1993;9:15-21.

________________________________________________________________________

Radiographically detectable lung cancer is hardly early disease. To be seen on routine chest radiograph, in the favorable circumstance of a peripheral nodule that does not overlie shadows of rib or mediastinal structures, a lesion has to be approximately 1 cm in diameter. Such a mass will typically contain 10 to the 9th tumor cells, representing about 30 doublings under an ideal condition of no cell loss. Such conditions never occur in human tumors; a comparison of actual and potential doubling times suggests that cell loss factors in the range of 80% to 90% are common.[3] At this point, the "early" tumor has undergone most of its life span. This long preclinical history for even the smallest radiographically detectable tumors gives ample opportunity for the mutational appearance and clonal selection of phenotypes capable of invasion, metastasis, and drug resistance.

The TNM staging system, while reasonably applicable to patients with non-small cell lung cancer (NSCLC), generally is not used for patients with small cell lung cancer (SCLC). The distinction between "limited" and "extensive" SCLC then reduces to a reflection of the sensitivity of the imaging and other technology used to search for it. The advent of computed tomography, magnetic resonance imaging, and polymerase chain reaction in staging means that there is less "limited" disease now than in the era of radionuclide imaging of these sites. These changes in stage distribution based on increasing sensitivity of staging need to be considered when we report apparent progress in treatment of subsets of patients.[4]

Shifts of stage distribution due to earlier diagnosis should not be confused with the shifts in staging classification.[5] In the case of changes in diagnostic sensitivity, clinicians are intervening in the disease process at an earlier time, which may result in either a real therapeutic gain (if effective therapies are available) or an apparent gain (from the lead time gained in time from diagnosis to time of death). In contrast, shifting patients from one stage to another by the use of more sensitive imaging techniques for staging but not changing the time of initial diagnosis simply changes the way in which the same set of clinical events are categorized and has no effect on the survival of the entire population with the disease, but may improve the outcome for each of its subgroups.

Figures taken from the clinical data on which the new International Staging System was based represent that which can be obtained with good standards of practice in staging and treatment.[6] They do not reflect the modest improvements in survival now being reported for the use of adjuvant chemotherapy in combination with either surgical resection or definitive radiation therapy for locally advanced disease.[7] In selected series with meticulous surgical staging of the mediastinum, survival for patients with T1N0M0 disease is up to 80% higher in the series of the Lung Cancer Study Group.[8] Survival for patients with T1N0M0 lung cancer is comparable to survival of patients with T1M0N0 breast cancer, but only a small percentage of patients with lung cancer are detected at this stage. It is a reasonable hypothesis that detection of patients with lung cancer prior to the development of nodal metastases would significantly improve survival. Patients with treated lung cancer who relapse generally do so within the first three years; true late failures are uncommon. Second primary tumors of the lung, esophagus, and head and neck sites are problematic, however, and in aggregate have an incidence of approximately 2% to 3% per year.[9] Thus, the impact of screening and early detection programs ought to be demonstrable with relatively short follow-up.

Lung Cancer Causation and Epidemiology

At the beginning of this century, lung cancer was a rare disease. The present global epidemic, with over 2 million deaths estimated for the year 2000, is the direct result of governmentally sanctioned production and aggressive marketing of addictive tobacco products, primarily cigarettes.[10] While an effective strategy for lung cancer treatment and control will include a broad spectrum of activities, it is unarguable that the greatest long-term reduction in lung cancer mortality will come from a decrease in the number smokers.

In the United States, lung cancer is most commonly diagnosed in the seventh decade of life. A generation ago, lung cancer was predominantly a disease of men. Much early clinical research in lung cancer, including the three large prospective trials of radiographic and cytologic screening conducted in the US, was limited to men who smoked. However, the increase in cigarette smoking by women starting in the 1940s changed this situation dramatically. Deaths from lung cancer overtook those of breast cancer for US women in the 1980s. Current US data show a decline in smoking prevalence among white men, but smoking has continued to increase among women, and recent data have suggested a leveling off or even reversal in the tendency to decreased smoking in young women.[1]

At present, approximately 25% of the adult population of the United States are smokers, and an additional 40 to 50 million are former smokers.[1] The American Cancer Society Cancer Prevention Study II has demonstrated that, while the risk of subsequent development of lung cancer declines for both men and women regardless of the age at which they quit smoking, the greatest gains are seen for those quitting at an earlier age, and that this difference is significant even when correcting for the number of years smoked.[2,11]

Lung Cancer Biology

The lung is anatomically and physiologically at least three separate organs. The trachea and main bronchi are normally lined by ciliated, pseudostratified columnar epithelium and also contain neuroendocrine cells. The predominant types of tumors arising in large central airways are squamous cell and small cell carcinomas. The thickness of the lining epithelium gradually declines as the airways become smaller. The pseudostratified, ciliated columnar cells gradually give way to ciliated columnar and finally ciliated cuboidal cells in the terminal bronchioles. Epithelial mucous cells are interspersed throughout the conducting airway. Interspersed among the cuboidal cells of the terminal and respiratory bronchioles are Clara cells, thought to produce the mucus covering for these small airways. The predominant histology seen in peripherally arising lung cancers is adenocarcinoma, which morphologically can be divided into solid and bronchoalveolar types. At the cellular level, these tumors arise from type II pneumocytes that normally produce surfactant. The bronchoalveolar carcinomas are believed to arise from Clara cells that are involved in xenobiotic metabolism. Each of these cell types has associated characteristic differentiation markers that may form the basis for both detection and therapeutic strategies (Table 2).[12]

Table 2. -- Markers of Lung Cancer Differentiation
__________________________________________________________________________

From Mulshine J, Linnoila RI, Treston AM, et al. Candidate biomarkers for application as intermediate end points of lung carcinogenesis. J Cell Biochem Suppl. 1992;Suppl 16G:183-186.
______________________________________________________________________________

Approximately one fifth of lung carcinomas are large cell undifferentiated tumors that cannot be assigned to one of the above lineages. All histologic types can be found admixed within a single tumor, consistent with a model of their development from a common stem cell of variable differentiation potential. The usual anatomic distribution of the different tumor histologies may derive from the normal distribution of partially committed cell lineages, from variable penetration of different carcinogenic components of cigarette smoke to different regions of the lung, and possibly from differences in local metabolic transformation of procarcinogens and effects of the extracellular matrix and paracrine growth factors on carcinogenesis.[12]

These distributions of normal cell and tumor types are typical rather than absolute. Lung cancers of all cell types may be found in any location in the tracheobronchial tree and lung. This classification implies that, while there may be some very early events common to the development of all types of lung cancer, further preneoplastic and neoplastic development can follow along several divergent lines, and screening strategies should be able to detect each of these. The observation that there has been a shift in the proportion of lung cancers of the various histologic types over the past several decades, with the predominant cell type changing from squamous cell to adenocarcinoma, should be considered in a proper theory of lung cancer initiation and promotion.[13]

Historical Screening Studies

In the 1950s, several nonrandomized trials of screening for lung cancer were conducted in the United States and Europe. Trials in Philadelphia[14] and London[15,16] used chest photofluorograms obtained at six-month intervals. These studies found that approximately one half of the cancers detected in these populations were found by the screening examinations, and the remainder was identified on interval chest radiographs obtained for evaluation of symptoms. While the resectability rate of the cases detected on the screening examinations was approximately 30%, the overall resectability for all cases was only 20%, which did not appear different from historical results in unscreened patients.

Table 3. ­ Prospective Trials of Screening for Lung Cancer
_______________________________________________________________________

_______________________________________________________________________

In an attempt to clarify questions raised by these nonrandomized studies, the US National Cancer Institute sponsored lung cancer screening trials that were conducted in the 1970s in three institutions -- Johns Hopkins University (The Johns Hopkins Lung Project [JHLP]), the Mayo Clinic (Mayo Lung Project [MLP]), and Memorial Sloan-Kettering Cancer Center (MSKLP).[17-19] In addition to these three US studies, a randomized trial has been conducted in Czechoslovakia,[20] and several recent carefully conducted case-control studies have taken place in Europe and Japan. The individual trials differed somewhat in their design and study population (Table 3).

The JHLP and the MSKLP trials were designed to evaluate the incremental benefit of adding sputum examination to chest radiography (CXR). In both the JHLP and MSKLP trials, the control groups were offered annual CXR. The dual screen group was offered an annual cytological examination of induced sputum plus examination of spontaneously produced sputum at four months and eight months. The comparison, therefore, was whether the addition of regular sputum cytology examinations to annual radiographic screening led to a reduction of lung cancer mortality compared with annual radiographic screening alone. Of the US studies, only the MLP trial was designed to compare unscreened management with a policy of intensive dual screening with both CXR and sputum cytology examinations every four months. The unscreened group, however, was advised to obtain annual CXR and sputum cytology as part of "routine medical care," but these individuals were not reminded to comply with this initial recommendation. Thus, this study can be seen as comparing two different frequencies of recommended screening and compliance.

The outcomes of these studies have been presented in several formats: results of the initial prevalence examinations in the cohorts to be screened, the follow-up of the patients detected during the screening period, and overall pooled results of the three trials. The general observations and conclusions of the three trials are clear, but their interpretation has been controversial. There was agreement that "within the trials there was no advantage in terms of mortality reduction to the group offered intensive screening in the Mayo study, with four monthly sputum cytology and chest radiology, while in the Johns Hopkins and Memorial Sloan-Kettering studies, there was no advantage to the group that received sputum cytology in addition to annual chest radiograph examinations."[21] However, the trials as designed did not address or resolve the question of whether any pattern of surveillance CXR was better for high-risk populations than a policy of obtaining such studies only when patients presented with symptoms.

These studies led to a conclusion that while the screening approaches available at that time could lead to the detection of presymptomatic lung cancer, particularly squamous cell carcinoma, at an earlier stage than in the control group, the overall survival for the screened population was no longer than that of the controls. This led to adoption of policies discouraging routine CXR and sputum screening and provided an unwarranted nihilism regarding all aspects of lung cancer detection and treatment.

None of the US randomized trials compared a policy of screening vs no intervention in the absence of symptoms. The Czechoslovak trial most closely approached this design; all enrolled men underwent an initial prevalence examination, after which the control group had no other planned intervention for three years. Screened individuals underwent CXR and cytology every six months. In the initial three years of the study, 36 lung cancers were detected in the study group vs 19 in the control group, and survival of these patients with lung cancer was superior in the intervention group compared with that in the control group, but the overall death rate from lung cancer was greater (although not significantly -- 28 vs 18 cases) in the study group (P=0.18). At three years, CXR and sputum cytology were performed for both groups, followed by annual CXR. At six-year follow-up, the mortality rate was the same for the two groups.

The studies of issues and populations in these trials, which were designed in the late 1960s and conducted in the early 1970s, are not entirely pertinent for the mid-1990s. The screened populations were all men, most still smokers, and often afflicted with other tobacco-related illnesses that led to a high frequency of interval CXRs in addition to the screening examinations. The present candidate for screening is more likely to be a man or woman 40 to 59 years of age, a former smoker with a history of 15 to 25 pack years, and often without illnesses requiring close medical attention. In addition to this shift in demographics, a change in the dominant histology of newly diagnosed lung cancer has emerged. During the period of the US collaborative trial, most cases in both the screened and unscreened groups were squamous cell carcinoma.[19,22] This pattern has changed in the past decade, for unclear reasons, to one in which adenocarcinoma is the predominant histology in trials of patients with both unresectable and resectable disease. In the past, adenocarcinomas of the lung have been more common in women, but they have now become the predominant histology for both sexes in the US.[13] European series continue to report higher incidence of squamous cell carcinoma.

Use of Molecular Markers in Screening

An ideal marker of genetic change should appear early in the carcinogenic process, yet be specific for malignancy or for a commitment to subsequent malignant development. It should be detectable in body tissues that are easily and repeatedly obtainable by relatively noninvasive and inexpensive procedures, such as exfoliated cells or peripheral blood rather than bronchoscopic biopsies. The genetic change should be readily detectable by automated or semiautomated methods, and structural changes in coding regions of a gene sequence may be better targets than changes that alter gene regulation. Finally, situations in which a limited number of genetic changes account for the majority of tumors are appealing in that they limit the number of mutant sequences to be screened.

For these reasons, genetic changes in the ras oncogene have been an appealing target. Mutational activation of ras occurs in approximately 30% of lung adenocarcinomas, as well as in high frequency in adenocarcinomas of the bowel and pancreas. Activation almost always involves point mutation at codons 12,13, or 61.[23] Presence of ras mutations in lung adenocarcinoma is closely linked with prior cigarette smoking and appears to be an adverse prognostic factor independent of stage.

Sidranski et al[24] first demonstrated that ras mutations could be identified in exfoliated cells present in the stool of patients with resectable and potentially curable colon cancer. Subsequently, Tobi et al[25] were able to detect ras mutations at codon 12 in exfoliated colonic mucosa in 40% of clinically normal individuals at high risk for developing colorectal cancer because of strong family history or a personal history of adenomas. Others have reported detection of mutant ras in gastric aspirates and stool specimens of patients with pancreatic cancer.[26,27]

Several recent reports extend these data to lung cancer. Kelly et al[28] have demonstrated the feasibility of detecting ras mutations in cells obtained by sputum cytology in patients with known lung cancer. Mills et al[29] have described using a sensitive assay for K-ras codon 12 mutations in individuals suspected of having lung carcinoma. They studied 87 specimens of bronchoalveolar lavage fluid that had been obtained from 86 patients, 35 of whom underwent subsequent diagnostic bronchoscopy. Of 52 patients with diagnosed lung cancer, 16 had bronchoalveolar lavage with K-ras 12 codon mutations, including 14 of 25 patients with adenocarcinoma, 1 of 3 with bronchoalveolar carcinoma, 1 of