In 2011, recommendations for a multidisciplinary classification of lung adenocarcinoma were published under the auspices of the International Association for the Study of Lung Cancer, the American Thoracic Society and the European Respiratory Society. The review was considered necessary due to emerging data on the radiological features, genetics and therapeutic approaches to lung adenocarcinoma, all underpinned by expanding the knowledge of the pathology of this common tumour. The existing WHO classification of 2004 was not really fit for this multidisciplinary focus on the disease.
This review describes the recommendations made on the reporting of surgically resected lung cancers according to their predominant pattern, and argues the case for replacing the term bronchioloalveolar carcinoma (WHO 1999 and 2004 definition) with adenocarcinoma in situ and for the introduction of minimally invasive adenocarcinoma. There is also a discussion of diagnosis of non-small-cell lung carcinomas in the small biopsy or cytology setting, a practice that was inadequately addressed in WHO 2004, yet this is much more relevant to most pathologists’ daily practice because 85% or so of adenocarcinomas are never resected. Predictive immunohistochemistry, used correctly, can reduce non-specific diagnosis to less than 10% of the cases. Finally, there is an overview of the emerging data on therapeutically relevant lung adenocarcinoma genetics, considering targetable mutations that are now the focus of much activity. The clinical relevance of these changes is discussed.
- Lung cancer
- Molecular pathology
Statistics from Altmetric.com
In 2011, a multidisciplinary panel of lung cancer specialists representing the International Association for the Study of Lung Cancer (IASLC), the American Thoracic Society (ATS) and the European Respiratory Society (ERS) had published new proposals for the multidisciplinary classification of lung adenocarcinoma,1 the culmination of more than 4 years of discussion, debate and literature review. Several factors were responsible for driving this project including continued confusion over the term bronchioloalveolar carcinoma (BAC), a better understanding of the precursor lesions of peripheral adenocarcinomas, emerging genetic data on lung adenocarcinomas, a need for better diagnosis on small diagnostic samples and the development of effective molecular-targeted drugs. The existing 2004 WHO classification is primarily a book for pathologists, is aimed primarily at diagnosis of surgically resected tumours, it has some unsatisfactory aspects to the adenocarcinoma classification and does not address many issues that is now relevant in the multidisciplinary management of lung cancer.2 It has lost much of its clinical relevance. This paper reviews the essential aspects of the new classification proposals, including the data published since February 2011, with an emphasis on the clinical relevance of the proposed changes. Readers are referred to the manuscript itself for more comprehensive details.
Most pulmonary adenocarcinomas probably arise from precursor lesions known as atypical adenomatous hyperplasia (AAH) by transforming into a lesion now known as adenocarcinoma in situ (AIS), though formerly known as BAC.3–5 The peripheral bronchioloalveolar epithelial compartment which gives rise to these lesions is often referred to as the terminal respiratory unit (TRU),6 one of the characteristics of which is the expression of thyroid transcription factor 1 (TTF1). About 75%–80% of surgically resected lung adenocarcinomas express TTF1 indicating this TRU lineage. There is debate about the origin of those remaining TTF1-negative lung adenocarcinomas. Some may be TRU derived but have lost TTF1 expression, TTF1 expression appears to diminish as tumours lose differentiation. Some may arise from larger bronchiolar or even bronchial epithelium. Precursor, in situ lesions are not well understood in this context,7 ,8 but such an origin, even as a de novo growth without a morphologically recognised precursor would not be expected to express TTF1.
AAH and AIS are characterised by the lesional cells lining alveolar walls, replacing normal bronchioloalveolar epithelium, but there is no loss of alveolar architecture. This is lepidic growth. The histological distinction between AAH and AIS can be difficult, reflecting the fact that AIS evolves from AAH, and is based on cell density, pleomorphism and cell size.9 As both lesions preserve alveolar structure and contain air, the larger lesions over 1 cm in diameter may be visible on high-resolution CT scans (HRCT) as localised ‘ground glass’ opacities (GGOs).10 ,11 The WHO definition of BAC is a lesion lacking evidence of invasion.2 Such lesions pose no metastatic risk12 and are thus much better regarded as AIS. Furthermore, given the widespread confusion and misuse of the term BAC, rehearsing former, historical meanings which no longer meet the criteria, including invasive adenocarcinomas with a BAC component or multifocal adenocarcinomas showing evidence of ‘alveolar wall spread’ or lepidic growth, the term BAC should be discontinued.
Invasion begins in AIS lesions usually accompanied by collapse of the alveolar architecture and, critically, neofibroplasia leading to a fibroblastic scar in association with invading tumour cells.13 ,14 There are no data to indicate what the risk of invasion may be and so the clinical significance of either AAH or AIS is uncertain. Usually, this is not an issue as both of these lesions are normally encountered in the context of a lung resection for invasive adenocarcinoma, this latter more advanced disease determining the patient's outcome. Studies have failed to demonstrate the convincing effect on patient prognosis when AAH lesions are found with resected adenocarcinoma, although the author has some data to suggest that large numbers of AAH lesions, or AIS lesions may predict a slightly poorer postoperative outcome.15–18 Follow-up of long-term survivors of adenocarcinoma resection with AAH lesions shows anecdotal evidence of second primary adenocarcinomas, implying an ‘adenocarcinoma diathesis’. HRCT performed in screening trials for lung cancer or for other indications may reveal small localised GGOs.19 Most of these are probably not AAH or AIS, but AIS in particular may be discovered in the absence of invasive disease. AIS tend to be larger than AAH and therefore more likely to be seen on HRCT images. The majority of AAH lesions are too small to be detected. The diagnosis of AIS cannot, however, be secured until the lesion is resected (table 1). This has led to some major diagnostic challenges in frozen section diagnosis.20 Rare cases have been observed over time, and invasion has developed after many years.21
If the invasive focus within a lesion, which is otherwise AIS and less than 3 cm in diameter, is less than 5 mm across, the patient still appears to have no metastatic risk. This important observation, of adenocarcinomas with better than expected prognosis, mandated the creation of a category of minimally invasive adenocarcinoma (MIA).12–14 The loss of alveolar architecture associated with the evolution of alveolar collapse and invasion distorts and condenses the GGO pattern on HRCT and is associated with a mixed solid or GGO lesion. Such an appearance on CT has a high predictive power for adenocarcinoma,22–24 but the designation of MIA requires microscopic examination. Surgical practice is evolving in this area but if some standard for sublobar anatomical resection becomes established for AIS or MIA, then any required intraoperative diagnosis will be very difficult. Early experience suggests that a diagnosis of MIA can be difficult, even on formalin-fixed, paraffin-embedded material but there is evidence that some diagnostic consistency can be achieved.25 ,26 Obviously, MIA cannot be diagnosed on small diagnostic samples (table 1).
Surgically resected invasive adenocarcinoma
At least in contemporary European practice, the diagnosis of MIA is rare. Most resected adenocarcinomas are established in invasive lesions with large areas of invasive disease. In the WHO classification, four patterns of adenocarcinoma (BAC, acinar, papillary and solid with mucin) are described, which define four diagnostic subtypes of tumour. A fifth subtype is the mixed category, where more than one of these four patterns is seen. It became apparent that the vast majority of resected adenocarcinomas were ‘mixed’ and the other four categories were almost redundant. Subsequent data emerged, demonstrating that the predominant pattern in a mixed adenocarcinoma had prognostic relevance.27 At the same time, a micropapillary pattern of adenocarcinoma was recognised as being important.28 ,29 Data were gathered on the prognostic influence of these five histological patterns of adenocarcinoma, and since publication of the recommendations, other studies have confirmed the strong prognostic effect of pattern subtype.30–32 These findings have been correlated with those of positron emission tomography scanning.33 Resected adenocarcinomas should be reported with a qualifying statement as to the predominant subtype (lepidic, acinar, papillary, micropapillary or solid) (figure 1 and table 1). Furthermore, it was recommended that pathologists should describe the individual components in a semiquantitative manner to the nearest 5%, the so-called comprehensive histological subtyping.34 Tumours that are predominantly lepidic pattern disease showed a relatively good postoperative prognosis, those which are predominantly solid or micropapillary showed poor postoperative outcomes, with acinar and papillary predominant cases in between. A large recent study, however, showed papillary predominant cases to behave as badly as solid or micropapillary predominant cases.32 The basis of this difference may be through differences in interpretation between lepidic and papillary disease.35 Studies have shown that inter-observer agreement varies according to the patterns in question but that education can improve outcomes.26 ,36
These proposals have not been universally welcomed, principally due to a lack of immediate clinical relevance. There are many factors that pathologists might identify in a tumour, which may be ‘good’ or ‘bad’ prognostic indicators; yet, apart from tumour, node and metastasis (TNM) stage, none of these is routinely used in clinical decision making. TNM stage is the only recognised basis for selecting postsurgical patients for adjuvant therapy but survival improvements are modest.37 This method of selection seems rather crude. Disease of stage 2a or higher in the resection specimen, the trigger for adjuvant chemotherapy, predicts a <70% 2 year postoperative survival. A similar postoperative survival is seen, in stage 1 patients, with predominantly micropapillary or solid-pattern adenocarcinomas; yet, this information is not currently acted upon. Trials are needed, but pathologists should continue to provide these refinements on their reports and promote the potential clinical value of these data.
Multiple foci of invasive adenocarcinoma are not uncommon in surgical resection specimens, even though the clinical suspicion was of a solitary tumour. The author estimates that 10%–15% of the resected adenocarcinoma cases show synchronous tumours; some quoted figures are higher.1 This has important implications for staging the disease, and therefore for adjuvant treatment. If two or more lesions are deemed ‘related’, one metastatic from the other, then T3 disease is present, as opposed to two separate T1 lesions. The use of comprehensive histological subtyping has been shown to be equally, if not more, effective than global genetic analysis, in determining whether tumours are related or synchronous primary tumours.34 There are clear biological differences between these scenarios, although actual clinical outcomes using this approach to make the distinction are few.
As will be discussed later, the identification of pharmacologically targetable genetic alterations in lung adenocarcinoma has revolutionised treatment of this disease. Histological pattern shows some weak association with some tumour patterns. Some studies show correlation between lepidic disease and EGFR mutation, ALK gene rearrangements are associated with solid or cribriform histology, often with mucinous signet-ring cells and BRAF mutations may be associated with micropapillary adenocarcinoma.38–40 Lepidic predominant mucinous adenocarcinomas, formerly called mucinous BAC, are strongly associated with KRAS mutations and never bear EGFR mutations.1 Using histological pattern would not be a robust approach for routine case triage for testing, but for example, if limited material for testing required test prioritisation, such associations might help in decision making. One exception may be the mucinous lepidic case; the author would not routinely seek EGFR mutation in such a tumour.
Small sample diagnosis of adenocarcinoma
Accurate subtyping of non-small-cell lung carcinomas (NSCLC) on small biopsy samples or on cytology is challenging in many cases. Actual hard diagnostically defining criteria for either squamous cell or adenocarcinoma are few and, in the case of adenocarcinoma, more dependent on architecture than on individual cellular features. Keratinisation and intercellular bridges are the only defining criteria for squamous cell carcinoma, and both of these are prone to overinterpretation. Limited material in small diagnostic samples may show an NSCLC, but no differentiated features to allow further classification.
As already mentioned, the WHO 2004 classification gives little guidance on the diagnosis of small samples and diagnostic criteria for several subtypes of NSCLC (including large cell carcinomas, sarcomatoid carcinomas and adenosquamous carcinoma) require the whole tumour to be examined before the diagnosis is made.41 Attempts to force small samples of tumour-lacking definite features into a WHO category led to substantial inaccuracy42–46 and spawned the recommendation to use the term NSCLC-not otherwise specified (NOS) when diagnostic features were absent.42 It is worth noting that, among wrongly ascribed cases, adenocarcinoma and large cell carcinoma (almost by definition) predominated. The term large cell carcinoma should not be used when small diagnostic samples show only undifferentiated tumour. Instead, the use of the term NSCLC-NOS was a perfectly sensible approach, improved accuracy when cases were subtyped (up to 85% accuracy), but left 25%–40% of the cases in this undifferentiated group.47 If resected, a majority of the cases called NSCLC-NOS on small samples, proved to be adenocarcinoma on resection.47
This situation was acceptable, up until the choice of chemotherapy for advanced NSCLC varied according to non-small-cell subtype. Licensed indications for the use of the monoclonal anti-vascular endothelial growth factor antibody bevacizumab excluded squamous cell carcinoma for risk of fatal haemorrhage.48 The antifolate pemetrexed shows superiority over gemcitabine, in combination with cisplatin, in the treatment of advanced stage adenocarcinoma and undifferentiated carcinoma.49 The reverse is true in squamous cell carcinoma, a fact reflected in the prescribing guidance for pemetrexed. As will be described later, choice of cases for molecular genetic testing is driven by NSCLC subtype. Accurate subtyping of NSCLC in patients with advanced disease and only small diagnostic samples available, is now of critical clinical importance.
This clinical problem of high NSCLC-NOS rates was an important target for the new adenocarcinoma recommendations. Predictive immunohistochemistry (IHC) is the key to reducing the NSCLC-NOS rate to less than 10% of the cases. Undifferentiated cells, from a tumour that is differentiated as squamous or adenocarcinoma in other parts not sampled, may express lineage markers associated with the differentiation type. To predict squamous cell carcinoma p63, cytokeratin (CK) 5/6 and p40 are the most frequently used markers, while for adenocarcinoma TTF1 is by far the best, although Napsin A or even CK7 also get used (figure 2). A mucin stain remains a useful, specific but somewhat insensitive marker for adenocarcinoma. Predictive markers are statistically associated with NSCLC cell type, but none is unique to, or defining of, a particular tumour type. Predictive accuracy is around 85%.50–53 The diagnostic value of these markers can only be truly assessed in studies where morphologically NSCLC-NOS cases are used, there is no case selection, and surgical resection is used to prove the correct diagnosis. Many published studies on this topic do not meet these criteria.
The IASLC/ATS/ERS adenocarcinoma recommendations suggest that a minimum approach is to use p63 and TTF1 in NSCLC-NOS cases.1 One study used p63, CK5/6, TTF1 and a mucin stain to reduce a 25% NSCLC-NOS rate to 6%, with 83% predictive accuracy in an unselected cohort of cases, all of which were subsequently resected.50 Other studies have shown a similar performance.51–53 This approach also works well on suitably prepared cytology samples allowing IHC to be performed.53 ,54 Some have recommended p40 instead of p63, citing its greater specificity with a reduced tendency to stain solid pattern adenocarcinomas.55 Whichever IHC solution is used, pathologists should test the predictive performance of their chosen IHC panel as preanalytical and other technical issues may vary outcomes. Use of IHC, or more importantly, of a tissue, should be minimised, given the importance of molecular analysis as the next diagnostic step after tumour typing. The overuse of unnecessary IHC wastes tissue, time and laboratory resource and is responsible for the inability to complete subsequent molecular analysis on a proportion of cases. Double staining IHC techniques may help preserve tissue.
It is also important to be aware of the amount of staining (intensity, distribution) that is predictive. In the author's experience, and assuming the context of NSCLC-NOS by morphology, TTF1 expression of any degree predicts adenocarcinoma, whereas p63 may be expressed at low levels in some poorly differentiated adenocarcinomas, but is only rarely seen in the extensive, moderate or strong pattern more predictive of squamous carcinoma. Such p63 expression in adenocarcinoma is never associated with CK5/6 expression, the latter being the rule in squamous cell carcinoma. CK7 is almost universally expressed in adenocarcinomas, but is not infrequent in squamous carcinomas, usually at a lower level but the author does not consider it as a useful discriminating marker. In the author's experience, Napsin A adds little to the predictive power of TTF1, an opinion shared by some,51 but not others.56
This predictive IHC approach can reduce the NOS rate to less than 10%, the best that can be achieved. NSCLC-NOS cannot be completely eliminated because around 10% of all NSCLC cases are completely undifferentiated large cell or sarcomatoid carcinomas; some but not all of these cases have a squamous or adenocarcinoma predictive immunophenotype, the significance of which is not known and does not change the morphological diagnosis. Alternative, expensive molecular profiling will not solve this issue. In recognition of the predictive nature of a diagnosis predicated on IHC, it is recommended that cases should be called NSCLC, ‘probably’ or ‘favour’ either squamous or adenocarcinoma as appropriate (table 1). Clinically, such cases should be managed as for their predictive diagnosis.
Molecular pathology of lung adenocarcinoma
Another key component of the IASLC/ATS/ERS adenocarcinoma recommendations is the integration of molecular features into the final, complete diagnosis.1 In the WHO 2004 classification, genetics features prominently in the text, but this is descriptive, concerning pathogenesis.41 The molecular pathology that is so important now was barely known at the time of writing. In our diagnosis of adenocarcinoma, we now have three steps: morphological diagnosis, IHC refinement where needed and molecular pathology. Molecular characterisation is less well developed for other lung tumour types but will evolve.
Adenocarcinomas are now frequently subdivided into small groups, according to what are considered as the key driver mutations for each group.57–60 Pie charts (figure 3) are commonly used to depict the fact that a common tumour, lung adenocarcinoma, is now divided into numerous, often rare, subgroups, in total accounting for up to two-thirds of the cases. These subdivisions have critical clinical importance as they define groups of patients with targetable molecular alterations in their tumours. To date, experience of tyrosine kinase inhibitors targeting mutant EGFR (such as erlotinib, gefitinib and afatinib) or ALK fusion proteins (crizotinib) indicates excellent clinical responses and marked improvement in progression-free survival. The lack of overall survival benefit may be related to treatment cross-over in trials, but these are still worthwhile clinical outcomes.61–63
These diagrams are potentially misleading. Although there is good evidence that in lung adenocarcinoma, KRAS, EGFR, BRAF, HER2 mutations and ALK rearrangements are generally mutually exclusive, MET amplification is not and the status of others is not known. Much of the published data come from tertiary referral centres and patient selection will probably increase the prevalence of these mutations in those institutions compared to a ‘real world’ unselected population. Testing methodology and policy also play a part. Some testing approaches lack sensitivity or range, while other methods may overestimate mutation prevalence. With the exception of KRAS mutation, and probably MET amplification, those other mutations shown in figure 3 are unrelated to tobacco carcinogenesis. While these mutations are much common in never-smokers with adenocarcinoma, there is no known reason why a tobacco smoker should be any less likely than a never-smoker to develop such a tumour. The causes or risk factors for these mutations are unknown. If adenocarcinoma testing is limited to non-smokers, up to half of EGFR mutations will be missed.64 The higher levels of EGFR mutation observed in adenocarcinomas in women and those of East Asian ethnicity at least in part reflects a population in which smoking is uncommon and therefore, adenocarcinomas with these driver mutations are less diluted by tobacco carcinogen induced cases, as may be seen in a European patient cohort.
Reference was made earlier to the association of some mutations with adenocarcinoma patterns. A question often arises around testing squamous cell carcinomas for these mutations that are found in adenocarcinomas. The literature is confused on this issue, as many reports of EGFR mutation and even ALK rearrangement concern cases diagnosed on small samples where accuracy is lower, especially when predictive IHC is not used. Virtually, all EGFR-mutated tumours are TTF1 positive, reflecting the role EGFR mutation can have in TRU adenocarcinogenesis. EGFR mutation has no role in central bronchial carcinogenesis through squamous dysplasia and carcinoma in situ.4 Most studies of surgically resected squamous cell carcinomas fail to demonstrate EGFR mutations, although exceptional cases do appear.65 ,66 Technical issues can explain some results but misclassification is the root of most of these cases.67 Very exceptional observations should not drive clinical practice; however, it is recommended that molecular testing is carried out in those rare instances of a peripherally located squamous cell carcinoma occurring in a never-smoker or long-time ex-smoker.68 Testing is justified in a case where cell type is dubious; it might be adenocarcinoma.
The current reality of molecular testing of adenocarcinoma, in most European centres at least, is a search for EGFR mutation and ALK rearrangements. Figure 3 indicates the potential for many other biomarker-defined therapeutic targets. The danger of having no tissue left for molecular testing, after overzealous IHC, has already been mentioned. This is a particular danger in adenocarcinoma when pathologists practise defensively and pursue possible metastatic disease in the lung from a range of other sites, when none of these possibilities is clinically likely, in order to save embarrassment in the multidisciplinary team (MDT) meeting. Having said this, many patients now survive breast, colorectal or renal carcinoma, only to present later with lung cancer. The need for much better communication between all members of the MDT, to ensure the correct investigations are carried out in patients with lung adenocarcinoma, is emphasised by the new classification recommendations. Equally, there is a strong obligation upon those obtaining tissue for diagnosis to maximise yields, while of course, maintaining patient safety. It is possible that next-generation sequencing methods may provide a multiplex platform allowing multiple genes to be tested in the same small sample. This is, however, not a current reality in most centres as it will pose many practical difficulties and will not solve all our biomarker needs. IHC is still likely to play an increasing role in this regard.
The new IASLC/ATS/ERS adenocarcinoma classification was written by a multidisciplinary group of lung cancer specialists, and with the MDT in mind, precisely because emerging data on the pathology, radiology, diagnosis and treatment of lung adenocarcinoma have important clinical implications for patients with this common tumour. These recommendations included research implications and follow-up work has been published.69 ,70 These issues are relevant to all specialists, not just pathologists, and to everyone who needs to be aware of the full spectrum of pathological, radiological and therapeutic issues, and not just those related to her/his own speciality.
New adenocarcinoma classification recommendations embrace developments in radiology, therapeutics and biological understanding of this disease, to provide more clinically relevant information to all members of the MDT.
Competing interests None.
Provenance and peer review Commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.