Context.—Recent advances in human imaging technologies reawakened interest in lung cancer screening. Although historic and current preliminary and noncontrolled studies have not shown a decrease in lung cancer mortality in screened populations, many explanations have been proffered while the lung cancer community awaits the results of several large controlled population studies.
Objective.—To critically review the current model of adenocarcinoma development against the background of lung cancer screening results combined with observational pathologic and radiographic studies.
Data Sources.—Published articles pertaining to lung cancer screening, lung adenocarcinoma pathology, and radiology accessible through PubMed form the basis for this review.
Conclusions.—The current adenocarcinogenesis model is probably valid for many but not all lung adenocarcinomas. Screening data combined with radiographic and pathologic studies suggest that not all lung adenocarcinomas are clinically aggressive, and it is uncertain whether all aggressive adenocarcinomas arise from identified precursors.
Given the aggressive nature of lung cancer and the general inability of the world to curtail the abuse of tobacco products, we now must confront the bitter truth that lung cancer is the deadliest human cancer worldwide and that current treatment options are just not adequate.
Recent medical advances in lung cancer care include noninvasive and minimally invasive methods of diagnosing and treating lung cancer and pharmacogenetic approaches to chemotherapy selection. However, technologic advances have truly reinvigorated lung cancer clinicians and researchers alike. Although chest x-ray screening for lung cancer was left for dead following the disappointing results of the Mayo Clinic (Rochester, Minnesota), Johns Hopkins (Baltimore, Maryland), and Memorial Sloan-Kettering (New York, New York) studies from the 1970s, high-resolution computed tomography (HRCT) screening has forced lung cancer specialists to reconsider their long-held assumptions about the disease. The past 10 years have witnessed an accelerated pace in the radiologic characterization of lung lesions and the histologic classification of lung adenocarcinoma.
Although the fierce and often acrimonious debate regarding the efficacy of HRCT screening of high-risk individuals continues—and its future as a public health policy is unknown—the dialectic has produced an original and thoughtful reexamination of our most basic accepted notions regarding lung adenocarcinoma. This review aims to present our evolving understanding or lack of understanding of lung adenocarcinoma within the context of lung cancer screening.
ADENOCARCINOMA OF THE LUNG AND LUNG CANCER SCREENING FINDINGS
Not unlike cancer models employed in other organ systems, the current lung adenocarcinoma model assumes a progression from a few malignant cells to a group of cells to a small “early-stage” carcinoma, leading to “advanced-stage” carcinoma and death. This model suggests that all lung adenocarcinomas are stage I before progressing to higher stages and causing death.1,2 Thus, early-stage and late-stage carcinomas are separated only by time and the molecular events that occur during this time. This long-accepted scheme appears reasonable and provides the framework for much research; however, lung cancer screening data do not convincingly support this general view.3 Lung cancer screening with early detection should lead to an increase in early-stage cancers, a decrease in late-stage cancers, and lower lung cancer mortality. Yet, published data do not show these results and suggest that perhaps not all clinically early lung adenocarcinomas progress to advanced lung cancers.
Randomized trials of chest radiography with or without sputum cytology performed in the 1970s noted a large increase in early-stage cancers in the screened populations but not a meaningful decrease in the number of advanced cancers.4–8 The Mayo Lung Project, London and Czechoslovakian studies, demonstrated a near doubling of the number of early cancers without a comparable decrease in numbers of advanced-stage cancers. Lung cancer mortality was not decreased, despite the early detection and subsequent treatments. Analysis of additional follow-up of the Mayo Lung Project and Czech studies continues to support these initial conclusions.9–11 Additionally, although the current model predicts a drop in the number of advanced cancers diagnosed between scheduled screening radiographs, this was not the case, at least in the Czech study.4 However, one should not lose sight of the fact that these chest radiograph and sputum studies detected mainly central squamous cell carcinomas rather than peripheral adenocarcinomas.
Nevertheless, HRCT screening studies to date confirm aspects of the chest radiography studies. Although HRCT detects almost 10 times the number of early lung cancers compared with control populations, this screening modality does not decrease the incidence of advanced lung cancers.12–18 In fact, in several studies a few additional advanced cancers were noted compared with expected values.19 Lung cancer mortality data are not yet available.
Epidemiologic data to date do not strongly endorse the current cancer model either. One would expect a valid model to feature similar risk factors for both early and advanced clinical lung cancer.20–22 But instead, as seen in one large cohort screened for lung cancer, the prevalence of screening-detected early carcinomas was higher in women than in men, despite less tobacco exposure for the women.20
Whether one believes that tumor growth rates are exponential or follow a Gompertzian model,23 it appears that HRCT-detected lung cancers often have doubling rates in excess of 400 days in comparison with aggressive lung cancers with doubling rates of less than 270 days.24,25 In fact, 25% of stage I lung carcinomas in a recent Mayo Clinic HRCT screening study had volume doubling times of at least 400 days.25 These very slow doubling times suggest that screening tests identify “indolent” lung carcinomas. This term indolent is not a great descriptor, and epidemiologists prefer to speak about “overdiagnosis.” Recent mammography screening results suggest that some breast cancers might spontaneously regress.26 Could one consider such a possibility for some lung adenocarcinomas?
Perhaps the precursors of aggressive lung cancer are not these slow-growing or stable incidental lesions. Lung cancer screening studies definitely identify small, aggressive carcinomas, and these patients have prolonged survival times but, possibly, not improved disease-specific mortality. These seemingly conflicting findings are ascribed to “lead-time bias” by epidemiologists.27 Maybe there are 2 forms of lung adenocarcinoma, not unlike the accepted natural-history model of prostate cancer, and just maybe lung cancer screening studies are unmasking this fact. A bipartite model of lung cancer development has been proposed recently.3 The pathology of small lung adenocarcinomas may shed some light on the dilemmas facing lung cancer screening investigators.
OBSERVATIONAL STUDIES OF THE PATHOLOGY OF LUNG ADENOCARCINOMA
Although much information can be gleaned from the 1970s chest radiograph studies performed in Europe and in the United States, more recent HRCT screening studies from the Far East, Europe, and North America have drastically altered our perception of peripheral lung carcinomas. These studies essentially describe the radiographic and morphologic features of clinically unsuspected lung cancers (Figure 1).
Much of our current thinking derives from the pre-HRCT screening era, and the rebirth of lung cancer screening studies must be viewed within this context. A 1995 seminal work demonstrated that peripheral lung adenocarcinomas 2.0 cm or smaller with a pure lepidic or bronchioloalveolar growth pattern and no stromal invasion had a 100% 5-year survival rate.28 This finding shook the lung cancer community and led to the search for factors responsible for the astounding outcome. During the past 10 years, several additional Japanese studies replicated these results, with an emphasis on the 100% 5-year survival rate.29–32 This concept was and remains so striking that the World Health Organization classification of lung tumors modified its description of bronchioloalveolar carcinoma (BAC) in 1999 (Table 1).33 The classification for the first time made a distinction between BAC and invasive adenocarcinomas. The definition of BAC was altered to create a unique tumor type with a pure lepidic growth pattern lacking stromal, vascular, or pleural invasion. In this scheme, BAC can broadly be considered in situ adenocarcinoma.
As an in situ carcinoma, should the lung pathology community replace the well-established but paradoxical term BAC with the more appropriate adenocarcinoma in situ? Such a change would be logical if BAC is indeed a precursor to invasive carcinoma and has no invasive or metastatic potential. Although the Japanese thoracic surgery community is correct in investigating the role of limited lung resections for BAC—because lobectomy might be “overtreatment”—might one also consider simply following these lesions before making an a priori assumption of malignant potential. Recall that the Western oncology community requires gastric and colonic adenocarcinomas to demonstrate invasion in addition to architectural and cytologic severity, whereas our Japanese counterparts consider gastric and colonic glandular high-grade dysplasia to be carcinoma.34,35
Given the rarity of BAC, our understanding about the origin of invasive lung adenocarcinomas must evolve. The 1995 publication noted that individuals with 2.0-cm or smaller adenocarcinomas composed of both BAC and invasive components had a 75% 5-year survival rate, whereas those with a purely invasive growth pattern had a 52% 5-year survival rate.28 Molecular studies also seem to suggest a progression from BAC to invasive adenocarcinoma.36 Yet, up to 20% of 1.0-cm as well as a greater percentage of slightly larger invasive adenocarcinomas do not present as stage I cancers and do not enjoy spectacular survival in well-regarded studies.37,38 Obviously, these carcinomas do not resemble BAC.
Surgical pathology publications during the past decade have explored morphologic characteristics of solitary, small, peripheral, invasive adenocarcinomas in the hopes of identifying histologic prognostic factors. Following the belief that solitary small BAC is in situ carcinoma, numerous studies with slightly different methodologies announced that a variety of light microscopic findings, including the size of the parenchymal scar, percentage of lepidic growth, percentage of papillary growth, presence of vascular invasion, size of invasive focus within BAC, and location of stromal invasion with respect to the scar and the lepidic pattern, hold prognostic significance and even confirm the current adenocarcinogenesis model.29–32,39 Molecular classifiers and cytokine gene expression signatures are also highly touted by investigators.40–43
But these retrospective morphologic and molecular prognostic studies should be viewed with great care.44 Discerning malignant glands entrapped by or enveloped in scar (ie, BAC/in situ carcinoma) from malignant glands invading stroma (ie, invasive adenocarcinoma) is not a reproducible exercise for even the most experienced surgical pathologist. In addition, most of these studies derive from Japanese patients and have not been convincingly reproduced in Western, including HRCT screening–detected, cohorts.45
On account of the revised BAC definition, the World Health Organization also reordered the remaining major adenocarcinoma categories (Table 1).33,46 Because invasive lung adenocarcinomas rarely have a single morphologic pattern, a mixed subtype was added to the acinar, papillary, bronchioloalveolar, and solid patterns. These carcinomas are the most common adenocarcinomas and feature varying percentages of acinar, BAC (lepidic), papillary, and solid patterns. High-resolution computed tomography screening studies found that 70% to 90% of the detected adenocarcinomas, including subcentimeter lesions, are of mixed subtype.25,45,47–49 Unfortunately, this morphologically diverse subset is doomed to remain a category of convenience rather than a true pathologic entity, and attempts to investigate the role of morphology in the natural history of these adenocarcinomas is next to impossible, because semiquantification of the percentage of each subtype in a tumor is a poor surrogate for precise classification.
Prognosticating on the basis of morphology is also nearly impossible, given the lack of a reproducible and meaningful adenocarcinoma grading scheme. Remarkably, lung adenocarcinomas, unlike breast, kidney, bladder, and virtually all other epithelial neoplasms, have no accepted prognostically significant criteria with which to grade tumors. Even the World Health Organization is mute on lung adenocarcinoma grading. Thus, one is not certain whether, for example, the papillary and solid subtypes of lung adenocarcinoma owe their presumed aggressive behavior to their architectural patterns, cytologic features, mitotic rates, or some yet unrecognized molecular fingerprint.50–54 Advanced genotyping studies have not to date offered any guidance. Thankfully, the International Association for the Study of Lung Cancer is embarking on an adenocarcinoma grading project.
Thus, renewed interest in lung cancer screening has really accelerated the need for answers in the lung adenocarcinoma field. Prognostication and evolution of tumors have become high-priority areas of research in the lung pathology community. Lung cancer imaging studies perhaps offer a framework for further investigations.
OBSERVATIONAL STUDIES OF THE RADIOLOGY OF LUNG ADENOCARCINOMA
Late 20th and early 21st century thoracic radiology studies describe the HRCT correlates of BAC and invasive adenocarcinomas. Although a variety of different terms have been used for the radiographic appearance of adenocarcinoma nodules, there are essentially 3 different CT patterns: (1) ground glass opacity (GGO), (2) mixed density or mixed GGO, and (3) solid. However, not all pure GGO lesions are neoplasms, let alone BAC.55 Nodular GGOs usually represent focal interstitial fibrosis, inflammation, hemorrhage, or incidental lesions, including minute meningothelial-like lesions or even, remarkably, atypical adenomatous hyperplasia (AAH).56–58 Also, radiologists are not all equal in their abilities to identify and qualify small GGOs or nodules.59 Nevertheless, radiologic-pathologic correlation studies suggest that a radiographic GGO pattern from an adenocarcinoma corresponds to a histologic BAC pattern, whereas radiographic solid components indicate the presence of a histologic invasive component.55 Lesional size, percentage of GGO, presence of alveolar collapse, coarse spiculation, air bronchograms, and thickening around bronchovascular bundles have been suggested to correlate with varying prognostic factors, including lymph node metastases, vascular invasion, or even simply prognosis.60–64 Unfortunately, no more than 38% of HRCT-detected lung adenocarcinomas initially manifest as GGOs. Lamentably, in one study 7% of 1.0-cm or smaller pure GGO pattern tumors featured invasive adenocarcinoma morphology, and almost 39% of pure GGOs measuring between 1.0 and 2.0 cm contained invasive adenocarcinoma.65
Serial HRCT studies describe the radiographic progression of lung adenocarcinomas with GGO components and are a tempting stand-in for pathology in the natural-history scheme.64,66–68 Carcinomas initially presenting as GGOs are reported to increase in size in three-quarters of cases and develop solid components in almost 20% of cases.66 Solid components may enlarge or collapse, may become denser, and spiculations often develop. These findings seem to mirror the microscopic findings observed in mixed-subtype adenocarcinomas presumably arising from BAC/in situ adenocarcinomas. But these retrospective radiographic observations lack true pathologic correlates. With reported localized GGO doubling times ranging from more than 1 year (438 days) to an astounding 3.25 years (1188 days), one is left to wonder what exactly one is following.24 Given what is known about the long lifespan of lung adenocarcinomas, these doubling times may represent a slow growth phase, as predicted by a Gompertzian model, or a truly stable lesion. Obviously, scanning and rescanning GGOs for months or years before histologic assessment offers no real insight into the qualitative changes of the lesions. This may be the most significant hurdle for HRCT lung cancer screening programs.
CURRENT VIEW OF LUNG ADENOCARCINOMA PROGRESSION
Although lung cancer screening studies question the natural-history model of lung adenocarcinoma, this skepticism does not suggest that the model is entirely invalid, but rather that there must be more to the story.
This peripheral adenocarcinogenesis model suggests an adenoma-carcinoma sequence of events.1,2 Incidental minute foci of glandular atypia (atypical adenomatous hyperplasia [AAH]) noted in lobectomy specimens harboring adenocarcinomas are very well studied, and much evidence suggests that at least some peripheral lung adenocarcinomas arise from these lesions. In fact, the World Health Organization considers AAH a putative precursor lesion and defines it as a usually less than 0.5-cm focus of mild to moderately atypical type II pneumocytes or Clara cells that line alveoli and sometimes respiratory bronchioles in the absence of significant underlying interstitial inflammation (Figure 2). Although identified in no more than 3% of the general population, at least 1 focus of AAH is identified in up to 20% of primary cancer-bearing surgical lung resections.69–75 Lesions are far more prevalent in lung adenocarcinoma–bearing lobectomies rather than squamous cell–bearing or large cell carcinoma–bearing lobectomies, but the true incidence and prevalence are probably underestimated. Morphologic observations, including peribronchiolar fibrosis and bronchiolar scars in AAH-rich resection specimens, also suggest that AAH may regress (oral communication with K. M. Kerr, FRCPath, December 2008). Interestingly, tobacco use is not associated with AAH, and a sex association has not been well studied.73 Morphometric, cytofluorometric, immunohistochemical, and genetic data all support the original investigators' claims that AAH is a preinvasive neoplastic proliferation similar to but distinct from BAC.76,77 Atypical adenomatous hyperplasia should probably be considered a dysplastic lesion that may or may not develop into BAC/in situ adenocarcinoma (Figure 3).
Yet, we do not know what percentage of AAHs become BACs, and we do not know what percentage of BACs progress to invasive adenocarcinomas (Figure 4). The radiology literature is not particularly helpful, as alluded to above, because there are no persuasive data regarding the percentage of pure GGOs that are stable, the percentage that progress to solid lesions, or the time it takes for those GGOs to progress to solid lesions. Pathologic findings of 509 lung tumors combined from several HRCT screening studies, where the median carcinoma size was 1.5 cm and visualized as either a GGO or mixed GGO, indicate that BAC represents only 10% of the diagnosed lung cancers, (almost all screening-detected adenocarcinomas are mixed-subtype cancers), and only very few cases featured at least 1 AAH lesion (Table 2).17,18,25,45,47–49 The implication is that most small lesions detected in screening studies are not the in situ carcinomas described in many retrospective Japanese studies.
The lack of a stronger association between AAH and BAC also suggests that the AAH cannot be the only adenocarcinoma precursor. Bronchiolar columnar cell dysplasia purportedly represents an adenocarcinoma precursor lesion, and atypical goblet cell hyperplasia is a suggested precursor of pediatric adenocarcinomas arising in congenital cystic adenomatoid malformations.78–80 Furthermore, endobronchial adenocarcinomas do not have recognized antecedent lesions (Figure 5). Cancer stem cell research proposes that particular lung stem cells may be precursors to airspace-derived carcinomas.81–84 Molecular studies of early pulmonary adenocarcinomas also suggest the existence of other pathways in addition to AAH-BAC–invasive adenocarcinoma.85 Perhaps unrecognized precursors represent incipient aggressive adenocarcinoma and are responsible for the poor clinical outcomes observed in individuals with small lung adenocarcinomas.
LUNG CANCER SCREENING AND MULTIPLE ADENOCARCINOMAS
For lung cancer screening to decrease lung cancer–specific mortality, deadly carcinomas must be identified in their curative stages. However, lung cancer screening studies identify multiple carcinomas in up to 25% of patients.17,25,45,47–49 Although it is well accepted that mucinous BAC is a diffuse process because of the intra-alveolar spread of tumor, perhaps a new lung cancer Weltanschauung would include the possibility that nonmucinous adenocarcinomas may also have “diffuse lung disease genotypes and phenotypes.” The pathology community has begun to address the practical ramifications of these HRCT findings.
Although historic studies noted that synchronous lung tumors had grim outcomes, most of the primary tumors were large central carcinomas. Screening studies have unequivocally shrunken the size of both primary and satellite lesions. Most of the multiple tumors reported in screening studies measure less than 3.0 cm. The current American Joint Committee on Cancer staging guidelines state that multiple synchronous tumors should be considered separate primary lung cancers unless lymphatic invasion is identified and the patient has extrapulmonary metastases. These criteria date from the mid 1970s and are based on the study of 7 synchronous squamous cell carcinomas and a single adenocarcinoma.86
Staging definitions of these lesions have varied over the course of the past few staging systems. Earlier versions of the American Joint Committee on Cancer staging system upstaged patients with T1 tumors found to have satellite lesions in the same lobe to T2, whereas additional nodules in another ipsilateral lobe qualified as T4. The current classification designates satellite tumor nodule(s) in the primary tumor-bearing lobe as T4 or in an ipsilateral nonprimary tumor-bearing lobe as M1. Suffice it to say that clinical studies do not support these high stage designations.87–91 The imminent seventh edition of the American Joint Committee on Cancer staging manual will place patients with satellite tumor nodule(s) in the primary tumor-bearing lobe in the T3 group or patients with an ipsilateral nonprimary tumor-bearing lobe in the T4 group, whereas only a contralateral lung nodule will be considered an M1a disease.92 How these modifications affect multifocal BAC remains to be seen. After all, whether multifocal BAC represents synchronous primary carcinomas or intrapulmonary metastases is unknown.
Molecular studies assessing the clonality of synchronous lung cancers offer a more precise method of discerning intrapulmonary metastases from synchronous primaries. The k-ras analyses seem most informative. Several recent studies demonstrated multifocal tumors with different k-ras mutation profiles, suggesting the independent nature of each lesion.49,93,94 Loss–of-heterozygosity studies demonstrate that synchronous, histologically similar adenocarcinomas of the lung represent a very heterogeneous group at the genetic level. Although molecularly homogeneous tumors most likely represent intrapulmonary metastases, the nature of molecular heterogeneous tumors is unclear because of the uncertainty as to whether heterogeneity is a consequence of multiple tumor clones or genetic instability continuing after metastatic spread of a primary single clone.
Lung cancer screening as a public health policy remains in “no-man's land” and will remain so at least until the completion of large, randomized, controlled studies in the United States and Europe. Yet, the pathology community has to date greatly benefited from the exercises. Screening data suggest that not all aggressive adenocarcinomas arise from noninvasive lesions. Although AAH is probably a precursor lesion to BAC/in situ adenocarcinoma, the overall significance of their roles in the development of invasive carcinoma is not certain. Radiographic studies have aided in describing the progression of at least some small adenocarcinomas, but both observational and molecular data suggest that the AAH-BAC invasive adenocarcinoma pathway is not the only pathway responsible for the development of aggressive carcinomas. The development of multiple synchronous adenocarcinomas is hardly explored or explained. Although nothing short of the complete destruction of the tobacco industry will drastically lower lung cancer death rates, continued work on the origins and development of lung adenocarcinoma in both smokers and nonsmokers should prove useful. If nothing more, the current state of the field provides a sound foundation for continued investigations into advanced imaging and biomarker studies, not to mention clinically applicable molecular testing of adenocarcinoma subsets.
The authors have no relevant financial interest in the products or companies described in this article.
Reprints: Douglas B. Flieder, MD, Department of Pathology, Fox Chase Cancer Center, 333 Cottman Ave, Philadelphia, PA 19111 (Douglas.Flieder@fccc.edu)