Morphologic Features of Adenocarcinoma of the Lung Predictive of Response to the Epidermal Growth Factor Receptor Kinase Inhibitors Erlotinib and Gefitinib

Abstract

Context

A subset of lung adenocarcinomas appears preferentially sensitive to epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs). EGFR-activating mutations and never smoking are associated with response to TKIs.

Objectives

To describe the morphology of adenocarcinomas responsive to TKIs, compare it to tumors in nonresponding patients, and correlate findings with EGFR mutations, gene copy number, and protein expression.

Design

Material from 52 EGFR TKI-treated patients was studied: 29 responders and 23 nonresponders. Adenocarcinoma subtypes and morphologic features were defined in histologic and cytologic material. EGFR mutations were detected by sequencing, copy number by chromogenic in situ hybridization, and expression by immunohistochemistry.

Results

Tumors from TKI responders tended to be better-differentiated adenocarcinomas with bronchioloalveolar carcinoma components. Nonresponders showed more heterogeneous morphology, higher grade, and more subtypes, and were more likely to show solid growth. In nonresponders, the only pure bronchioloalveolar carcinoma was mucinous, a subtype known to be negative for EGFR mutations. Using World Health Organization criteria, all tumors in both groups other than pure bronchioloalveolar carcinomas would be classified as adenocarcinomas, mixed subtype, thereby obscuring some of these distinctions. EGFR mutations were significantly more common in responders (22/29 vs 0/23; P < .001). Immunohistochemistry and chromogenic in situ hybridization results were not significantly correlated with EGFR mutations or response to TKIs in this study.

Conclusions

Overall, histologic differences exist between tumors that respond to TKIs and those that do not, although sampling affects classification, and there is significant histologic overlap between the 2 groups. Response is strongly associated with EGFR mutations.

Lung cancer remains the leading cause of cancer death in the United States and worldwide in both men and women.¹ The treatment of non–small cell lung cancer (NSCLC) has recently expanded beyond surgery and conventional chemotherapy with the identification of tumors responsive to tyrosine kinase inhibitors (TKIs). Histopathology, along with careful clinical/molecular correlation, plays a critical role in the selection of patients most likely to respond to these new small molecule agents.

We have previously shown² that among NSCLCs, adenocarcinoma, and, interestingly, the subtype adenocarcinoma with bronchioloalveolar carcinoma (BAC) features, appear preferentially susceptible to the effects of the TKI gefitinib (Iressa, AstraZeneca, Wilmington, Del). A similar agent, erlotinib (Tarceva, Genentech, South San Francisco, Calif), has also been used with success against the same spectrum of lung carcinomas. Adenocarcinoma with BAC features corresponds to the World Health Organization (WHO) classification of adenocarcinoma, mixed subtype with a BAC component.

The epidermal growth factor receptor (EGFR) is a member of a family of transmembrane glycoproteins that includes HER-2, HER-3, and HER-4 that functions as a receptor tyrosine kinase and is present on and is commonly overexpressed in most NSCLCs. Overexpression in NSCLC ranges from about 40% to 90%. Differences in expression may relate to differences in assessment techniques, definition of level of overexpression, and differences in the study populations.³ Agents that interfere with phosphorylation of critical tyrosine residues can block signal transduction through EGFR. Gefitinib and erlotinib are such agents. The US Food and Drug Administration approved gefitinib in 2003 as targeted monotherapy for the treatment of patients with locally advanced or metastatic NSCLC. Use of this drug was subsequently restricted when a trial in unselected lung cancer patients failed to show a benefit.⁴ Erlotinib received approval in 2004 and is commonly used in NSCLC patients.

We undertook the characterization of the morphologic features of lung adenocarcinomas in patients who showed response to gefitinib and erlotinib by examining tissues obtained for histologic or cytologic diagnosis prior to initiation of therapy. Our aim was to identify pathologic features among adenocarcinomas that are predictive of response to these agents. In light of the recently reported identification of specific activating mutations in the EGFR TK domain underlying responsiveness of NSCLC to gefitinib and erlotinib,^5–7 the recognition of morphologic features predictive of treatment response might lead to better understanding of the interaction of tumor histology and genotype, and might result in more successful targeted therapy.

MATERIALS AND METHODS

Case Selection

We studied tumor material from 52 patients treated with erlotinib or gefitinib on institutional review board–approved protocols. These included 29 cases of adenocarcinoma of the lung from patients treated with gefitinib (n = 18) or erlotinib (n = 11) who showed response according to Response Evaluation Criteria In Solid Tumors⁸ and who had cytologic or histologic material available for review. Briefly, Response Evaluation Criteria In Solid Tumors defines partial response in tumor shrinkage as at least a 30% decrease in the sum of the longest diameter of the target lesions. The Response Evaluation Criteria In Solid Tumors guidelines were developed to derive response rates from unidimensional measurement of tumor lesions instead of the usual bidimensional approach. The selection criteria for patients receiving gefitinib included any NSCLC histology and locally advanced or metastatic disease. Erlotinib was given under a similar clinical trial where patients were preselected on the basis of histology, which had to include BAC, BAC with focal invasion, or adenocarcinoma with BAC features. On review, however, several patients (Table 1, Nos. 28 and 29) originally classified with some component of BAC were reclassified as adenocarcinoma, acinar type. Case 29 was pleural fluid, classified as acinar carcinoma. The tumors in these patients were compared with the tumors in a group of 23 patients treated during the same period who had adenocarcinoma that failed to respond to gefitinib (n = 7) or erlotinib (n = 16). Only adenocarcinomas were included in the comparison group because the aim of the study was to identify pathologic features among adenocarcinomas that are predictive of response to erlotinib or gefitinib.² This was based on the finding that BAC histology and smoking history predicted sensitivity to gefitinib. Responses to these agents in other NSCLCs (eg, pure squamous cell carcinoma, large cell carcinoma, large cell neuroendocrine carcinoma) are now known to be extremely rare. Importantly, none of the cases in either group were selected on the basis of EGFR mutational status, chromogenic in situ hybridization (CISH), or immunohistochemistry (IHC) results.

Table 1.

Responders^*

Patient	Sex	Age at Diagnosis, y	Drug	Smoking Status	Tumor Type	Specimen	EGFR Mutation	EGFR IHC	EGFR CISH
1	M	41	G	Former	Acinar, BAC	Lobe	Exon 19 del	2	NA
2	F	62	G	Current	Adenosquamous	Lobe	None	2	3.7
3	M	67	G	Never	Acinar, BAC	Wedge	Exon 19 del	3	4.9
4	F	71	G	Never	Acinar, BAC	Wedge	None	3	12
5	F	74	G	Former	Acinar, BAC, papillary	Lobe	L858R	NA	NA
6	M	46	G	Current	Acinar, BAC	Lobe	Exon 19 del	NA	NA
7	F	73	G	Never	Acinar, papillary, BAC	Lobe	Exon 19 del	2	4
8	M	47	G	Never	Acinar, papillary, BAC	Lobe	Exon 19 del	0	5
9	M	60	G	Former	Acinar, papillary	Lobe	Exon 19 del	NA	NA
10	F	70	G	Never	Acinar, BAC	Lobe	Exon 19 del	3	3.6
11	F	69	G	Never	Acinar, papillary, BAC	Lobe	Exon 19 del	3	3
12	M	66	E	Former	BAC, papillary	Lobe	Exon 19 del	0	3
13	F	68	E	Never	Acinar, papillary, BAC	Lobe	L858R	NA	NA
14	M	57	E	Former	Acinar, papillary, BAC	Wedge	L858R	0	6.8
15	F	73	E	Never	Acinar, BAC	Lobe	R776C, L858R	0	2.3
16	M	86	G	Former	Acinar, BAC	Wedge	None	0	4.4
17	F	43	E	Former	Acinar, BAC	Wedge	None	3	4.9
18	F	48	G	Never	Acinar, BAC	Biopsy	None	NA	NA
19	F	63	G	Former	BAC, papillary	Biopsy	L858R	2	2.5
20	M	71	G	Never	Acinar	LN mets	Exon 19 del	3	4.4
21	F	48	G	Never	Acinar, papillary, solid	LN mets	L858R	3	21.4
22	F	53	G	Former	Acinar	Biopsy	Exon 19 del	NA	NA
23	F	63	G	Former	Acinar, BAC	Biopsy	None	NA	NA
24	M	58	E	Former	Acinar, papillary, BAC	Biopsy	Exon 19 del	0	7
25	F	44	E	Never	BAC	Biopsy	Exon 19 del	NA	NA
26	M	73	E	Never	Acinar, BAC	Biopsy	Exon 19 del	NA	NA
27	F	50	E	Never	Acinar	Pleural fluid	L858R	0	4.5
28	F	82	E	Former	BAC	Biopsy	None	NA	NA
29	M	74	E	Former	Acinar	Biopsy	L858R	3	3.2

^*EGFR indicates epidermal growth factor receptor; IHC, immunohistochemistry; CISH, chromogenic in situ hybridization; G, gefitinib; BAC, bronchioloalveolar carcinoma; E, erlotinib; NA, not available; and LN mets, metastasis to lymph nodes.

We include 2 cases (Table 1, No. 29 and Table 2, No. 23) with only cytologic material. Although a complete histologic classification is not possible with such limited material, we include them in this study because they are representative of the type of material submitted for analysis and because they demonstrate the ability to detect mutations in limited samples.

Table 2.

Nonresponders^*

Patient	Age, y/ Sex†	Drug	Smoking Status	Tumor Type	Specimen	EGFR Mutation	EGFR IHC	EGFR CISH
1	59/F	G	Former	Acinar, solid with necrosis	Lung	None	3	3
2	51/F	G	Never	Acinar, papillary, BAC	Lobe	None	3	21
3	55/M	G	Former	Acinar, papillary, BAC, micropapillary	Lobe	None	2	6
4	66/F	G	Current	BAC, acinar, papillary, necrosis	Lobe	None	1	10.4
5	68/M	G	Former	Acinar	Lobe	None	2	2
6	49/F	E	Former	Papillary, micropillary	Lobe	None	0	3
7	53/M	E	Former	Acinar	Lobe	None	0	2
8	56/M	E	Former	Micropapillary, acinar, solid	Lobe	None	0	3.3
9	72/F	E	Current	Acinar	Lobe	None	0	3.4
10	67/M	E	Former	Mucin, papillary	Lobe	None	0	2.1
11	62/F	E	Former	Acinar, BAC	Lobe	None	0	3.3
12	72/F	E	Former	Acinar, BAC	Lobe	None	2	2
13	75/F	E	Former	Acinar, papillary, BAC	Lobe	None	3	3
14	59/F	E	Current	Acinar, papillary, BAC, micropapillary	Wedge	None	0	2.4
15	69/F	E	Former	Acinar, BAC, papillary	Wedge	None	0	3.5
16	71/M	E	Former	Acinar, micropapillary, papillary, BAC	Wedge	None	0	2.5
17	78/M	E	Former	Papillary, mucin, micropapillary	Wedge	None	0	2.1
18	60/M	E	Former	Acinar, BAC	Wedge	None	3	2.3
19	69/F	G	Former	Poorly differentiated adenocarcinoma	Biopsy	None	0	6.8
20	48/F	G	Never	Acinar with mucin	Biopsy	None	3	12
21	71/F	E	Former	Acinar, solid with necrosis	Biopsy	None	0	3.4
22	33/F	E	Never	BAC (mucinous)	Biopsy	None	0	2.3
23	60/M	E	Current	Acinar	FNA lung	None	0	3

^*EGFR indicates epidermal growth factor receptor; IHC, immunohistochemistry; CISH, chromogenic in situ hybridization; G, gefitinib; BAC, bronchioloalveolar carcinoma; E, erlotinib; and FNA, fine-needle aspiration.

^†Age at time of diagnosis.

Pathologic Specimens and Criteria

The responder group specimens included 12 lobectomies, 5 wedge resections, 9 lung biopsies, 1 pleural fluid cell block, and 2 cases of metastatic adenocarcinoma to lymph nodes. The nonresponder group included 1 pneumonectomy, 12 lobectomies, 5 wedge resections, 3 lung biopsies, 1 lymph node biopsy with metastatic disease, and 1 cytology cellblock from a lung fine-needle aspiration biopsy. The number of histologic sections examined varied greatly, with up to 10 slides available for review on some cases, and others having only a single cell block section. This was due to the diversity of sample types (lobectomy, wedge resections, etc), grossing techniques, tumor size, and the fact that some of the cases were received in consultation and only representative slides were available. All cases were reviewed by M.F.Z. and S.H.

The tumors were studied for histologic subtype and the presence of a BAC component. The tumors were initially classified on tissue as pure BAC, mucinous, nonmucinous, or mixed; BAC with invasion (BAC-I) where the invasive component of the tumor was less than 15%; adenocarcinoma with BAC features when an adenocarcinoma with a prominent (at least 15%) BAC growth pattern was present; and adenocarcinoma without evidence of BAC. The categories of BAC-I and adenocarcinoma with BAC features correspond to the WHO nomenclature of adenocarcinoma, mixed subtype. Our categorization was consistent with the simplified schema we previously proposed; it allows for some quantification of BAC and recognizes a form of minimally invasive BAC.^2,9 We recognize that, as stated in the recent BAC pathology consensus paper,¹⁰ more studies are needed to define a “minimally invasive” category of BAC. To more completely define the histologic features of these tumors, all adenocarcinomas were retyped according to WHO criteria. As the size of the specimens varied from pneumonectomy to cytologic material, the possibility that the histologic features present were not representative of the entire tumor morphology is recognized. Nonetheless, we included incompletely sampled cases in our study, as they reflect the kind of material likely to be encountered in actual practice.

EGFR Mutation Analysis

Sequencing of EGFR exons 18 through 21 was done as described in a prior report.⁵ Briefly, genomic DNA was extracted from tumors embedded in paraffin blocks. All sequencing reactions were performed in both forward and reverse directions, and all mutations were confirmed by polymerase chain reaction amplification of an independent DNA isolate. For drug-sensitive tumors that did not have mutations, sequences from exons 19 and 21 were also determined from at least 2 independently derived polymerase chain reaction products. EGFR mutational analysis on a subset of these patients was previously reported.⁵ In some cases, screening for hotspot EGFR mutations in EGFR exon 19 (in-frame deletions) and exon 21 (L858R) was performed by nonsequencing polymerase chain reaction–based methods described in detail elsewhere.¹¹

Microdissection of tumor from the paraffin blocks was not performed, but in some cases gross trimming away of nontumor tissue was done to ensure that there was always more than 50% tumor in the material submitted for DNA extraction.

EGFR Immunohistochemistry

Immunohistochemistry for EGFR was performed using a monoclonal EGFR antibody (clone 31G7, Zymed Laboratories Inc, South San Francisco, Calif) according to the manufacturer’s instructions. EGFR results were interpreted as follows: 0, no membrane staining; 1+, faint, partial membrane staining; 2+, weak complete membrane staining in more than 10% of cancer cells; and 3+, intense complete membrane staining in more than 10% of cancer cells. For overall scoring, 0 and 1+ were considered negative, and 2+ and 3+ were counted as positive.

EGFR Copy Number Analysis

Determination of EGFR gene copy number was performed using an EGFR CISH kit according to the manufacturer’s instructions (Zymed Laboratories). Briefly, the sections were incubated at 55°C overnight. The slides were deparaffinized in xylene and graded ethanols. Heat pretreatment was carried out in the pretreatment buffer at 98°C to 100°C for 15 minutes. The tissue was digested with pepsin for 10 minutes at room temperature. After application of SpotLight digoxigenin-labeled EGFR probe (Zymed Laboratories), the slides were cover-slipped and the edges sealed with rubber cement. The slides were heated at 95°C for 5 minutes, followed by overnight incubation at 37°C using a moisturized chamber. Posthybridization wash was performed the next day and followed by immunodetection using the CISH polymer detection kit (Zymed Laboratories). The CISH signals were counted in at least 30 nuclei with a light microscope using a ×40 objective. A tumor was interpreted as positive for gene amplification when the average number of gene copies was 5 or more per nucleus.

Survival benefit was not analyzed in this study.

Statistical Analyses

The statistical analyses were performed using Fisher exact test. All reported P values are 2-tailed.

RESULTS

Clinical Features

Among the 29 responders, there were 12 men and 17 women. Fifteen responders were former or current smokers, and 14 were never-smokers. Ten of the women were never-smokers. The age range was from 41 to 86 years, and the average age was 62 years. Two of the patients, 14 and 17, were Asian women, and the remainder were white. Features of responders are summarized in Table 1. Neither Asian patient was found to harbor an EGFR mutation. EGFR mutations are found more frequently in the Asian population, but they are not ubiquitous. The tissues studied in both these cases were from wedge resections, so false-negative results from sampling were probably not responsible for the findings. Of the 23 nonresponders, 9 were men and 14 were women. A total of 20 were former or current smokers and 3 never-smokers; all 3 never-smokers were women. The age range was 33 to 78 years, with an average of 61 years. All patients were white. Features of nonresponders are summarized in Table 2.

Statistical comparison of the clinical features of the 2 groups showed that the proportion of never-smokers was the only significant difference (P = .02), being higher in the responder group.

Pathologic Features of Responders

Histology

All of the responders had adenocarcinoma, including 1 adenosquamous carcinoma. Both components of the adenosquamous carcinoma were sequenced together. Of the 28 histologic lung specimens, including 9 small biopsies, 22 (78%) had BAC (Figure 1) or a BAC component. The single cytology pleural fluid specimen was an acinar adenocarcinoma, and the 2 cases of metastasis to lymph nodes (patients 18 and 19) showed mixed subtype adenocarcinoma with papillary, and a small solid pattern in one of the nodes and an acinar pattern in the other. Micropapillary architecture, commonly found in nonresponders, was not seen (Figure 2). None of the tumors were mucinous. Necrosis was not seen. The tumors were initially classified into the categories of BAC, BAC-I, and adenocarcinoma with BAC features as described above. These tumors were also classified according to WHO terminology and almost all (26) of the cases fell under the mixed subtype category (Figure 3). No discernible morphologic differences were found between the responders treated with erlotinib and those treated with gefitinib.

Figure 1.

A, An area of bronchioloalveolar carcinoma (BAC) in a responder. This tumor was from a lobectomy specimen (hematoxylin-eosin, original magnification ×40). B, A focus of adenocarcinoma interpreted as BAC in a small biopsy specimen from a responder. The inset shows the intranuclear inclusions frequently found in BAC (hematoxylin-eosin, original magnifications ×4 and ×100 [inset]).

Figure 2.

A focus of micropapillary carcinoma seen in a nonresponder. This morphology was not seen in responders (hematoxylin-eosin, original magnification ×100).

Figure 3.

Mixed subtype adenocarcinoma can be found in both responders and nonresponders. This is from a nonresponder who underwent a wedge resection (hematoxylin-eosin, original magnification ×40).

Cytology

One of the specimens was from a pleural fluid cell block, and morphologic features consistent with acinar carcinoma were seen. Cytologic features associated with BAC were not identified in this specimen.

EGFR Status in Responders

A total of 22 (75%) of 29 responders had EGFR mutations. One case (No. 15) demonstrated 2 different EGFR mutations. Among the 7 responders whose tumors did not contain a mutation of EGFR, the histologies included BAC, acinar, acinar with BAC, and adenosquamous carcinoma. The EGFR copy number by CISH was increased in 5 (28%) of 18, and EGFR IHC was positive in 12 (63%) of 19 (Figures 4 and 5).

Figure 4.

Increased epidermal growth factor receptor (EGFR) copy number by chromogenic in situ hybridization in a responder. EGFR copy number was high (more than 6), and the specimen was from a wedge resection (original magnification ×40).

Figure 5.

Epidermal growth factor receptor immunohistochemical stain scored as 3+ in a responder. The specimen was from a lobectomy (original magnification ×40).

When the cases were separated into 2 groups—one that contained major resections (wedges and larger specimens) and one that contained small biopsies and cytology specimens—the results were similar. Mutations were found in 13 (76%) of 17 large resections and 9 (75%) of 12 small biopsy/cytology responders. EGFR IHC was positive in 8 (61%) of 13 large resections and 4 (66%) of 6 small biopsy/cytology specimens. A total of 3 (25%) of 12 larger specimens showed amplification by CISH, whereas 2 (33%) of 6 limited specimens were amplified (Figure 6). KRAS mutational status was unavailable.

Figure 6.

Chromogenic in situ hybridization showing epidermal growth factor receptor amplification in case 3, a gefitinib nonresponder. The copy number in this case was 6 (original magnification ×100).

Pathologic Features of Nonresponders

Histology

The nonresponders also showed heterogeneous adenocarcinoma, but in this group more subtypes per case were identified. As with the group of responders, the tumors were initially classified as BAC, BAC-I, and adenocarcinoma with BAC features, and then reclassified. Specifically, of the 23 gefitinib and erlotinib nonresponders, 1 was pure mucinous BAC, 3 more had mucinous adenocarcinomas (Figure 7), and 18 showed moderately to poorly differentiated adenocarcinoma with some amounts of mixed histology subtypes, including BAC features (amount of the BAC component ranging from very focal to more than 15%), papillary (Figure 8) and micropapillary carcinoma, and 1 was a poorly differentiated adenocarcinoma. More detailed histologic descriptions are provided in Table 2. Necrosis was seen in 3 cases (Figure 9). A solid growth pattern was seen in 3 cases. This is in contrast to the responders, where necrosis was not seen and a solid growth pattern was identified (metastatic to lymph node) in only 1 case.

Figure 7.

Mucinous bronchioloalveolar carcinoma (BAC) (case 22) from an erlotinib nonresponder. This tumor was homogeneous in appearance. Areas of nonmucinous BAC were not identified (hematoxylin-eosin, original magnification ×40).

Figure 8.

Papillary component in adenocarcinoma from patient 13, an erlotinib nonresponder (hematoxylin-eosin, original magnification ×40).

Figure 9.

Necrosis seen in patient 1, a gefitinib nonresponder. Necrosis was seen in 3 nonresponders (hematoxylin-eosin, original magnification ×40).

Cytology

The only cytology specimen in the nonresponder group was a fine-needle aspiration of the lung where the tumor was considered to be an acinar adenocarcinoma.

EGFR Status in Nonresponders

None of the 23 tumors in this group tested positive for EGFR mutations, a highly significant difference relative to responders (P < .001). EGFR CISH and IHC data were also available in all 23 of the nonresponders. The EGFR copy number by CISH was increased in 5 (22%) of 23, and EGFR IHC was positive in 8 (35%) of 23. Thus, there was no significant association of response to erlotinib and gefitinib with either EGFR expression or copy number in this series. KRAS mutational status was unavailable.

The larger specimen group of nonresponders had EGFR expression by IHC in 8 (44%) of 18 cases, and the limited biopsy/cytology group had 1 (20%) of 5 cases expressing EGFR by IHC. A total of 3 (16%) of 18 large resections had amplification by CISH, and 2 (40%) of 5 biopsy/cytology cases were amplified.

Correlations Between EGFR Parameters

There were no statistically significant relationships between EGFR mutational status and either amplification by CISH or EGFR expression by IHC. There was also no statistically significant relationship between EGFR positivity by IHC and EGFR copy number by CISH in the cases with data on both parameters.

COMMENT

Our results show that there are morphologic differences in the adenocarcinomas that respond to the EGFR kinase inhibitors erlotinib and gefitinib and those that do not. Although there is some overlap in the 2 groups, the responders tended to be better differentiated, more commonly had a BAC component, and showed less histologic heterogeneity than the nonresponders. There was no necrosis and very little solid growth pattern among responders. One responder showed an adenosquamous pattern of growth, and although a mutation was not found in this specimen, reports of EGFR mutations in adenosquamous carcinoma show the mutation to be present in both components of this tumor.¹² Although responses were seen in pure nonmucinous BAC, mucinous BAC or any mucinous adenocarcinoma did not show response to either gefitinib or erlotinib. Unlike other types of BAC, the mucinous variant of BAC is known to often harbor KRAS mutations,^13,14 which have recently been associated with primary resistance to EGFR kinase inhibitor therapy; KRAS mutations are mutually exclusive with EGFR mutations.¹⁵

Although pure BAC is rare, a BAC component may be found in up to 20% of adenocarcinomas in general. In our responder group, well to moderately differentiated adenocarcinoma and adenocarcinoma with BAC features (specifically acinar and BAC mixed subtype by the 2004 World Health Organization classification) predominated. This is consistent with other reports correlating histology and response to drug.^16–19 In addition, the most frequent subtype of adenocarcinoma to show EGFR mutations in our series was adenocarcinoma with BAC features, specifically mixed subtype with acinar and BAC components. Because the nature of the material available for review varied greatly, from fine-needle aspiration specimens to a pneumonectomy, and did not always include the entire tumor, there are limitations in the histologic classifications in this study group. Additional studies analyzing all material from resected tumors in a standardized fashion are needed to confirm our observations. Histologic classifications of tumors, including that of the WHO, are not based on cytologic specimens, and we recognize the inherent problems in the use of such specimens for this purpose. There are, however, well-documented features associated with BAC that can be seen in cytologic material,^20–22 and we believe that BAC features should be reported, when present, in cytologic samples originating from lung tumors.

There were more women than men in both the responders and nonresponders, perhaps reflecting the rising incidence of adenocarcinoma in women. There were a significantly greater number of never-smokers versus current or former smokers in the group of responders compared with nonresponders, consistent with other reports.² The observation that response to gefitinib and erlotinib and EGFR mutation status is both associated with smoking status and sex is well documented.^{2,5,18,19,23,24} Two patients who responded to EGFR kinase inhibitors were Asian, both with adenocarcinomas with a BAC component but, interestingly, these 2 tumors did not demonstrate EGFR mutations. There were no Asian patients among the nonresponders. Adenocarcinomas arising in Asians are known to be 2 to 3 times more likely to harbor EGFR mutations.^4,25

Our findings that most tumors with EGFR mutations had BAC features differ from some other reports comparing histology and EGFR mutational status. Shigematsu et al²⁶ reported no association of BAC features with EGFR mutation status, although their study included pure BAC tumors in the category of adenocarcinoma with BAC features, and they did not specify whether the BACs were mucinous or nonmucinous. Their report also did not comment on the amount of BAC present in the specimens, and no tumor photomicrographs were provided.

Kim et al²⁷ describe a significant association between the response to gefitinib and the dominant papillary subtype of adenocarcinoma. This is in contrast to our findings, where the papillary histology was more common in the nonresponder group. They divided their tumors into the categories of BAC, acinar, papillary, and solid. They did not describe a mixed adenocarcinoma and BAC subtype. Bronchioloalveolar carcinoma can have simple papillary structures within intact alveolar spaces²⁸ but should, according to the WHO definition, be excluded from the category of papillary carcinoma. The similarities between these 2 subtypes could lead to misclassification of papillary adenocarcinoma as BAC or vice versa. However, the photomicrograph included in their report convincingly demonstrates papillary adenocarcinoma.

Yatabe et al²⁹ found that adenocarcinoma with EGFR mutations had characteristic morphologic features, and they hypothesize that these tumors arise from the terminal respiratory unit (TRU) and propose the term TRU carcinoma. The authors state that TRU carcinoma may be difficult to recognize, and they define this entity as including most nonmucinous BACs, mixed acinar and BAC subtypes, and some papillary subtypes. They describe thyroid transcription factor 1 immunoreactivity as characteristic of TRU carcinoma. We did not examine the expression of this marker in our series. Although we did not attempt to classify our tumors using the criteria of Yatabe et al, it appears likely that most tumors with EGFR mutations in their series are morphologically similar to those with EGFR mutations in our study.

A total of 24% (7/29) of tumors responsive to erlotinib or gefitinib in our series demonstrated no EGFR mutation. These included 1 nonmucinous BAC, 5 mixed acinar and BAC, and 1 adenosquamous. The types of specimens in these cases were, respectively: a biopsy, 3 wedge resections and 2 biopsies, and 1 biopsy for the adenosquamous case. The failure to detect EGFR mutations in this group may have been due to the presence of less than 10% tumor cells in the sample submitted for molecular analysis (although there was no histologic evidence for this), or it may reflect the presence of as-yet undetected mutations in other tyrosine kinases sensitive to gefitinib or erlotinib.

No EGFR mutations were identified in the group of 23 nonresponders. All tumors were adenocarcinoma, including 1 poorly differentiated adenocarcinoma. These tumors were of higher grade than the responders and demonstrated more heterogeneity, consistent with previous reports.^17,19,30

Interestingly, there was no significant difference in the detection of EGFR mutations, positivity for EGFR by IHC, or amplification by CISH when the tumors were analyzed by size of the tissue sample. Although small samples may not show all histologic components present in a tumor, this finding indicates that cytology and small biopsy samples can be used for EGFR mutation analyses, particularly when metastatic or inoperable disease is present. Our series is, to our knowledge, the only one to systematically examine the morphologic features of lung adenocarcinomas that fail to respond to erlotinib or gefitinib and that lack EGFR mutations. Others have documented the lack of response in squamous carcinoma or large cell carcinoma,^3,16,31 neuroendocrine carcinomas, and in a variety of nonpulmonary cancers, including breast, colorectal, and prostate.

EGFR is commonly expressed in NSCLCs, particularly squamous cell carcinoma.²² It is well documented, however, that EGFR expression by IHC does not reliably predict response to EGFR-targeted therapies.^31–33 This is in agreement with our findings where 8 of the nonresponders showed EGFR immunoreactivity, and 5 of these showed intense labeling (3+). Of 19 responders studied, 7 failed to show any EGFR immunostaining. Nonetheless, because 12 (63%) of 19 responders had some EGFR IHC positivity, EGFR IHC, in conjunction with histologic features, may be helpful in selecting patients for EGFR TKI therapy if no other EGFR test is available. However, we found no statistically significant correlation between EGFR mutation status and overexpression of EGFR by IHC. Of the 23 nonresponders studied by IHC for EGFR, positivity was found in 8. No association between response to gefitinib and EGFR protein expression using the Dako antibody was found in the IDEAL trials.^34,35 Two smaller studies from Italy and Korea also could not verify a predictive role for the detection of EGFR protein expression.⁴

Several studies have used fluorescent in situ hybridization analysis to determine EGFR copy number in lung cancers. Suzuki et al³⁶ found that overexpression of EGFR in NSCLC was accompanied, in most cases, by gene amplification. Hirsch et al³² found frequent EGFR overexpression in NSCLC, especially in squamous cell carcinoma, and showed that it was correlated with increased gene copy number by fluorescent in situ hybridization. More recently, larger studies have found correlations between response to EGFR kinase inhibitors and EGFR copy number by fluorescent in situ hybridization (as well as similar but weaker correlations with EGFR expression level by IHC).^3,33,36 We found no statistically significant correlation between EGFR mutation status and amplification of EGFR detected by CISH, although a larger series may be needed to observe such correlations.

CONCLUSIONS

Although subtle, there are histologic differences between adenocarcinomas responsive to the EGFR kinase inhibitors erlotinib or gefitinib and those that show primary resistance to these drugs. Response is most strongly associated with the presence of the EGFR mutations. In general, responders tended to have better differentiated adenocarcinomas with significant amounts of BAC present and with less histologic heterogeneity. Necrosis and solid patterns of growth were less frequent in the responders. Papillary subtype was more frequent in the nonresponders, as was mucinous BAC.

Correlations between response to EGFR kinase inhibitors or the presence of EGFR mutations, and either EGFR overexpression by IHC or EGFR amplification detected by CISH, were not evident in this study of 52 patients. In the absence of detectable EGFR mutations or when mutational data are not available, histologic subtype may be useful in selecting patients with adenocarcinoma of the lung for treatment with EGFR kinase inhibitors.