EGFR mutations detected on cytology samples by a centralized laboratory reliably predict response to gefitinib in non–small cell lung carcinoma patients


  • See related editorial on pages 540–3, this issue.

  • We are grateful to Antonio Rossi (Division of Medical Oncology, “S. G. Moscati” Hospital, Avellino, Italy) for critically reading the article and to Jean Ann Gilder (Scientific Communication SRL, Naples, Italy) for text editing.



Epidermal growth factor receptor (EGFR) mutations are reliably detected by referral laboratories, even if most lung cancer cytology specimens sent to such laboratories contain very few cells. However, EGFR mutations may be distributed heterogeneously within tumors, thereby raising concerns that mutations detected on cytology are not representative of the entire tumor and, thus, are less reliable in predicting response to tyrosine kinase inhibitor (TKI) treatment than mutations detected on histology. To address this issue, the authors reviewed their clinical practice archives and compared the outcome of TKI treatment among patients who were selected by cytology versus patients who were selected by histology.


From July 2010 to July 2012, 364 cytology samples and 318 histology samples were received. Exon 19 deletions and the L858R point mutation in exon 21, detected by fragment assay and TaqMan assay, respectively, were confirmed by direct sequencing; discrepancies were resolved by cloning polymerase chain reaction products. The response rate (RR) and progression-free survival (PFS) at 12 months (range, 3-34 months) were evaluable in 13 EGFR-mutated patients who were selected for treatment by cytology and 13 patients who were selected by histology.


The mutation rate was similar in histology samples (8.5%) and cytology samples (8.8%). The RR (54%) and PFS (9.2 months) were similar in histologically selected patients and cytologically selected patients (RR, 62%; PFS, 8.6 months; P = .88). The disease control rate (responsive plus stable disease) was 92% in histologically selected patients and 100% in cytologically selected patients.


EGFR mutations detected on cytology specimens by a centralized laboratory can predict TKI treatment response equally well as mutations identified on histology samples. Cancer (Cancer Cytopathol) 2013;121:552–560. © 2013 American Cancer Society.


Epidermal growth factor receptor (EGFR) mutation analysis is part of the current standard of care in advanced non–small cell lung cancer (NSCLC).[1] EGFR is targeted by the tyrosine kinase (TK) inhibitors (TKIs) erlotinib and gefitinib.[2] These drugs occupy the TK adenosine triphosphate (ATP) binding site thereby preventing EGFR activation and downstream signaling effects. Gefitinib has been authorized in Europe as monotherapy in the first-line setting for NSCLCs harboring EGFR mutations.[3] These NSCLCs are more commonly associated with females, adenocarcinoma (ADC) morphology, and Asian ethnicity in the absence of a history of smoking.[2] Patients carrying in-frame deletions in exon 19 and the L858R point mutation in exon 21 have a high response rate (RR) to treatment and longer progression-free survival (PFS) than those who receive standard chemotherapy.[4]

Surgical treatment is not recommended for patients with advanced-stage NSCLC, and EGFR mutations are often tested on cytologic specimens.[5] Whereas, in the United States, there are many laboratories performing mutational tests in-house, in Italy, only a few pathology departments were equipped to run molecular diagnostics at the time EGFR testing became mandatory for the prescription of gefitinib in patients with NSCLC, who received various types of samples from outside hospitals.[5-8] Because the primary cytopathologist is often reluctant “to sacrifice” the morphology of malignant cells for DNA extraction, the smear sent to centralized laboratories is not always the “good one” and is often paucicellular.[6] In an attempt to overcome the problem of paucicellular samples, we recently applied a laser-capture microdissection (LCM) procedure that limits the number of cells required to detect EGFR mutations to as few as 25.[9] Savic et al,[10] Molina-Vila et al,[11] and, more recently, Chowdhuri et al[12] reported similar results, which demonstrated that, from a technical viewpoint, EGFR mutations can be reliably detected even on paucicellular lung cancer cytology specimens.

However, it is still debated whether EGFR mutations are distributed homogeneously within tumors or whether they coexist with wild-type (WT) neoplastic cells.[13, 14] This raises concerns that the mutation status assessed on paucicellular samples may not represent the mutation status of the entire tumor; therefore, it cannot predict the response to treatment. Cytology testing predicted favorable treatment outcomes in the small cohorts of patients investigated by Oshita et al[15] and Lozano et al.[16] However, the mutations were detected mainly on cellular smears from cytologic material that was collected in-house; hence, cytopathologists were able to select better quality smears. This does not reflect the current Italian practice, which is external centralized testing that is usually done on samples that contain very few cells.

To date, the reliability of predicting treatment outcomes based on EGFR mutational diagnosis performed by a centralized laboratory on cytologic material is still unknown. To address this issue, we reviewed our records of the clinical outcomes of patients affected by lung cancer and compared the benefit of treatment in 2 groups of patients: 1 selected for treatment by testing histology and 1 selected by cytology using the same assay. Our data indicate that EGFR mutations detected on paucicellular cytology specimens, such as those tested by a centralized laboratory, can predict TKI treatment response equally well as mutations identified on histology samples.


Patients and Samples

The Molecular Pathology Laboratory at the University of Naples Federico II is a centralized laboratory for EGFR testing for South Italy. Our laboratory conformed to the Italian Society of Pathology external quality 2011 audit with a genotyping score of 100% (available at: [accessed June 5, 2013]). From July 2010 to July 2012, we received 682 consecutive requests for EGFR mutation analysis relative to patients with lung cancer who had TNM stage IIIB or IV disease and ranged in age from 29 to 86 years (mean age, 59 years). After obtaining the patient's consent, oncologists and the primary pathologists from 13 institutions recorded the clinical and pathologic data (including the original pathology report) on a dedicated website. Then, the corresponding tissue sample was express-mailed to the centralized laboratory. Upon receipt of a sample, the information relative to light microscopy and immunocytochemistry included in the original pathology report was standardized according to the International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society classification (see Table 1).[17] Briefly, cases in which ancillary stainings had been crucial for tumor subtyping were classified as “NSCLC favor adenocarcinoma” or “NSCLC favor squamous cell carcinoma (SCC).” The most common antibodies in pathology reports were TTF-1 (clone 8G7G3/1; Dako, Glostrup, Denmark) as an ADC marker, and p63 (clone 4A4; Ventana Medical Systems, Inc., Oro Valley, Ariz) as a squamous marker. Smoking status was defined as follows: smokers were those who had smoked >100 cigarettes per lifetime, and never-smokers were those who had smoked <100 cigarettes per lifetime.

Table 1. Sex, Smoking History, and Non–small Cell Lung Cancer Subtyping of Histologic and Cytologic Specimens
CharacteristicNo. of Patients 
  1. Abbreviations: ADC, adenocarcinoma; NOS, not otherwise specified; NSCLC, non–small cell lung carcinoma; SCC, squamous cell carcinoma.

Tumor type    
NSCLC favor ADC513383 
NSCLC favor SCC639 
NOS  74 

A single tumor sample, either histologic or cytologic, was tested for each patient. Cytologic samples (n = 364) or histologic samples (n = 318) were received in similar proportions (53% vs 47%). Sample features and diagnoses are detailed below (see Histologic and Cytologic Specimens). Briefly, histologic samples were either surgical specimens (n = 52) or small biopsies (n = 266), whereas cytologic samples included 274 fine-needle aspirates, 31 transbronchial needle aspirations, 30 effusions, 27 bronchial washings/brushings, and 2 sputum specimens.

Histologic and Cytologic Specimens

Histologic samples were taken from a primary tumor (n = 245) or from metastatic sites (pleura, n = 32; central nervous system, n = 13; lymph node, n = 12; bone, n = 7; liver, n = 5; skin, n = 2; mediastinum [adrenal gland, n = 1; orbit, n = 1]). Fifty-two were surgical specimens, and 266 were small biopsies. The latter were obtained by bronchoscopy (n = 181), computerized tomography-guided core-needle (n = 39), cerebral stereotaxic (n = 12), pleural (n = 32), and punch (n = 2) biopsies.

Histopathologic reports were standardized according to the International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society classification. Straight morphologic diagnoses were as follows: ADC (n = 213), SCC (n = 24), in situ ADC (formerly bronchioloalveolar carcinoma; n = 3), adenosquamous carcinoma (n = 4), small cell lung cancer (n = 1), sarcomatoid pleomorphic carcinoma (n = 4), neuroendocrine (NEC) (n = 2), and mixed ADC-NEC (n = 1). NSCLC was not classifiable by morphology alone in 66 patients (20.7%). This diagnosis was refined by immunohistochemistry into NSCLC-favor ADC (n = 35), NSCLC-favor SCC (n = 2), and NSCLC-favor NEC (n = 4). In 26 patients (8.1%), further subtyping was not possible, and NSCLC NOS was the final diagnosis. Although ADC morphology was recognizable in 53 of 73 patients with metastatic disease (72.6%), immunohistochemistry was performed in all to confirm the pulmonary origin.

Fine-needle aspirates (n = 305) had been obtained in 271 patients by computerized tomography (lung, n = 261; liver, n = 5; bone, n = 4; central nervous system, n = 1), in 31 patients by transbronchial needle aspiration on mediastinal lymph nodes, and in 3 patients (subcutaneous, n = 2; thyroid, n = 1) on palpable masses. Exfoliative samples included effusions (n = 30), bronchial washings/brushings (n = 27), and sputum (n = 2). In total, 292 cytologic samples were taken from primary tumors, and 72 were taken from metastatic sites.

Information retrieved from cytopathologic reports revealed ADC (n = 227), SCC (n = 21), mixed ADC–SCC (n = 2), others (n = 6), and NSCLC (n = 108). In 57 instances, the latter were further subtyped by immunohistochemistry as favor ADC (n = 51) and favor SCC (n = 6).

Preanalytic Sample Processing

Assay validation

Before clinical implementation, we evaluated the lowest limit of detection (LOD) of our laboratory-developed assay in PC9 cells, which harbor a Glu746-Ala750 deletion in exon 19; in H1975 cells, which carry the L858R point mutation; and in A549 cells, which carry the WT EGFR gene. The A549 and H1975 cell lines were obtained from American Type of Culture Collection (Rockville, Md). The PC9 cell line was obtained from the National Research Council/Institute of Experimental Endocrinology and Oncology (Naples, Italy). To determine the LOD, we serially mixed the mutated (PC9 and H1975) and WT (A549) EGFR cell lines at dilutions of 50%, 30%, 20%, 10%, and 5%. The detection limits of exon 19 mutations and of the L858R point mutation were 10% (Fig. 1A) and 5% (Fig. 1B), respectively. Thus, a sample would need to be constituted by 20% tumor cells to reach the threshold of 10% mutant alleles of exon 19, assuming the mutation is heterozygous without amplification. To determine the minimal number of neoplastic cells needed to detect mutations, we laser-microdissected, Papanicolaou-stained slides that contained 5, 25, 50, or 100 PC-9 cells and H1975 cells.[9] Mutations were consistently detected on 100, 50, and 25 microdissected cells. The results were not consistent when only 5 cells were microdissected (data not shown).

Figure 1.

These charts illustrate evaluation of the lowest limit of detection (LOD) of mutation detection assays in serially diluted, epidermal growth factor receptor (EGFR)-mutated PC9 and H1975 cell lines and in a wild-type (WT) cell line (A549). (a) In an EGFR exon 19 fragment assay, peaks (green dots) reveal that a deleted allele is present in the 50%, 30%, 20%, and 10% dilution experiments. The 5% dilution and the WT control lacked mutant peaks. (b) The exon 21 TaqMan assay (Invitrogen Corporation, Carlsbad, Calif) detected the L858R mutation in the homozygous mutated control (blue dots) and in all dilutions tested (green dots), including the lowest (5%). Red dots indicate the WT control. There was no amplification of the nontemplate control (NTC). All samples in b are shown in triplicate.

On the basis of assay validation data, 2 pathologists (C.B. and G.T.) reviewed each specimen before testing and considered samples “satisfactory” if they contained at least 25 neoplastic cells or a larger number of cells of which >20% were neoplastic. In histologic specimens, 4 (resection specimens) or 5 (biopsy specimens) 5-μm-thick serial sections were obtained. The first and last sections were stained with hematoxylin and eosin (H&E) to verify tumor representativeness. In specimens that contained <20% neoplastic cells, tumor cell enrichment was performed using LCM. Preanalytic processing of cytologic samples differed according to the percentage of neoplastic cells and the type of preparation, because some laboratories use smears, whereas others use cell blocks. Smears were first photographed (DMD 108; Leica, Milan, Italy) to record the cytomorphology of a representative microscopic field for our archives. Then, the coverslips were removed by incubating in xylene at room temperature in separate 50-mL tubes to avoid contamination. Smears with >20% neoplastic cells did not require tumor cell enrichment, and tissue material was scraped from the whole slide surface with a scalpel blade and directly collected into tubes. Similarly, in the case of neoplastic-rich cell blocks, 6 to 9 paraffin curls also were collected without tumor cell enrichment. Conversely, in the case of samples that contained <20% neoplastic cells, the areas with the most neoplastic cells were marked either on a reference H&E-stained section (cell blocks) or with a diamond needle on the back of the slide (smears). Areas >2 mm in greatest dimension were manually macrodissected. Areas <2 mm were sampled by LCM to select a population of at least 25 neoplastic cells, as previously described.[9] A sample was considered satisfactory for molecular testing if mutations were present in exon 19 or 21 or if both were WT. A sample was considered unsatisfactory if polymerase chain reaction (PCR) amplification failed in exon 19 or 21 (absent a mutation in the other EGFR exon).

EGFR mutation analysis of exons 19 and 21

Genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Crawley, West Sussex, United Kingdom) according to the manufacturer's instructions. DNA was resuspended in 20 μL molecular biology water. We measured DNA quantity using the NanoDrop 1000 Spectrophotometer (Thermo Scientific, Milan, Italy).

Exon 19 EGFR deletions were detected by a fragment assay, and the exon 21 L858R mutation was detected using a TaqMan assay (Life Technologies, Monza, Italy). The primers, probes, PCR conditions, and criteria used to interpret results are reported elsewhere.[11] PCR products with an aberrant length fragment on electropherograms (exon 19) or an allelic discrimination plot (exon 21) were further processed by direct sequencing to verify the patient's mutational status and to type the exon 19 deletion. Direct sequencing was carried out as previously described.[9] In case of discrepant results or lack of confirmation by direct sequencing, PCR products were subcloned into a TOPO TA cloning vector (Life Technologies) according to the manufacturer's instructions. For each sample, 30 plasmids were purified and sequenced using the BigDye Terminator kit (version 3.1; Life Technologies) and run on the ABI 3730 analyzer (Life Technologies) with M13 forward and reverse primers. Sequence data were analyzed using a Mutation Surveyor (SoftGenetics, State College, Pa). A sample was scored as a true-positive EGFR mutation when the alteration was identified in at least 1 clone.[18]

Evaluation of Treatment Outcome

To determine the clinical effect of TKIs, we selected patients according to the following criteria: 1) evidence of an EGFR mutation on exon 19 or 21, 2) received gefitinib during the course of the disease, and 3) the availability of at least 3 months of follow-up. Twenty-six patients met these inclusion criteria (15 men and 11 women; median age, 54 years). The therapeutic effect of the TKI in patients who were selected by EGFR testing on cytologic specimens (n = 13) was compared with that in patients who were selected by testing histology (n = 13). Patients received gefitinib as first-line therapy (n = 16), second-line therapy (n = 6), or further lines of therapy (n = 4) during the course of the disease. Patients who were selected by histology had received gefitinib as first-line therapy (n = 7), second-line therapy (n = 3), and further lines of therapy (n = 3). Patients who were selected by cytology had received gefitinib as first-line therapy (n = 9), second-line therapy (n = 3), and further lines of therapy (n = 1). The median follow-up was 12 months (range, 3-34 months). Disease status was evaluated in all patients by total body computed tomography scan. The RR was evaluated based on Response Evaluation Criteria in Solid Tumors (RECIST) guidelines.[19] PFS was calculated from the start of gefitinib treatment to the date of confirmed progression or death from any cause. PFS data were plotted as Kaplan-Meier curves. P values ≤ .05 were considered statistically significant. All statistical analyses were performed with the IBM SPSS Statistics 18 software package (SPSS Inc., Chicago, Ill).


EGFR Mutation Rates

Overall, 599 of 682 samples (87.8%) met the criteria for adequacy and qualified for testing, including 294 histologic samples and 305 cytologic samples. Rejection rates were 7.5% (n = 24 of 318 samples) and 16.2% (59 of 364 samples) for histology (all unsatisfactory specimens were biopsies) and cytology, respectively. The sampling method for each of the 59 rejected cytologic samples is detailed in Table 2. PCR amplification failed in 1 histologic biopsy. Overall, 52 of 598 samples (8.7%) harbored EGFR mutations (34 exon 19 deletions and 18 L858R mutations). In 2 samples, the exon 19 deletion, which was detected by fragment assay and confirmed by PCR product cloning, corresponded to a WT sequencing electropherogram (Fig. 2). Mutations were more frequent in women (32 of 599 samples; P = .0001) and in nonsmokers (42 of 599 samples; P = .03), as detailed in Table 3. Mutations also were associated with ADC (48 of 599 samples; P = .007) diagnosed by light microscopy alone (80.7%; 42 of 52 samples) or suggested by immunocytochemistry (NSCLC-favor ADC; 11.5%; 6 of 52 samples). Three samples with mutations (5.7%) were diagnosed as NSCLC-not otherwise specified (NOS), and 1 was associated with a mutated NSCLC-favor SCC (1.9%).

Table 2. Sampling Method and Rejected Cytologic Specimens
Specimen TypeNo. TestedNo. RejectedTotal No.Percentage Rejected
  1. Abbreviations: FNAB, fine-needle aspiration biopsy; TBNA, transbronchial needle aspiration.

Pleuric effusion2463020
Bronchial washings/brushings2072724.13
Figure 2.

A partial response by a patient to gefitinib is illustrated, indicating discordance between fragment assay and direct sequencing. The 2 patients who had discordant results had a partial response to gefitinib treatment. (a) In a paucicellular adenocarcinoma cytologic smear, (b) the exon 19 deletion was detected by fragment assay and was missed by direct sequencing (results not shown). The patient had a partial response to gefitinib treatment, as indicated on computerized tomography scans (c) before and (d) after treatment.

Table 3. Mutation Distribution According to Sex, Smoking Status, and Non–small Cell Lung Carcinoma Subtyping
CharacteristicMutant Rate: No./Total No. (%)
  1. Abbreviations: ADC, adenocarcinoma; NOS, not otherwise specified; NSCLC, non–small cell lung carcinoma; SCC, squamous cell carcinoma.

Cytology, n = 29427/52 (51.9)
Histology, n = 30425/52 (48.1)
Women32/52 (61.5)
Men20/52 (38.4)
Yes10/52 (19.2)
No42/52 (80.7)
Tumor type 
ADC42/52 (80.7)
SCC1/52 (1.9)
NSCLC favor ADC6/52 (11.5)
NSCLC favor SCC0/52 (0)
NSCLC-NOS3/52 (3.7)
Other0/52 (0)

EGFR Mutation According to Sample Type

Sex (P = .12), smoking history (P = .5), and ADC histology (P = .08) did not differ between histologic and cytologic specimens, indicating that there were no statistically significant differences in the populations diagnosed by histology and cytology (Table 1). The NOS category was more frequent among cytologic than histologic specimens (14% vs 7%:P = .003). Mutation rates were similar in histologic samples (8.5%; 25 of 293) and cytologic samples (8.8%; 27 of 305). The histology and cytology groups had a similar distribution of EGFR mutations according to sex (P = .52), smoking history (P = .31), and NSCLC subtyping (P = .08).

Outcome of Gefitinib Treatment

Most patients responded to gefitinib treatment regardless of whether they were selected by histology or cytology. In total, 15 patients (7 selected by histology and 8 selected by cytology) had a partial response to gefitinib treatment according to RECIST guidelines,[19] including the 2 patients whose mutations were detected by fragment assay and were missed by direct sequencing (Fig. 2). Stable disease was obtained in 10 of 26 patients. Only 1 patient who was tested on histology had disease progression as the best response. The disease control rate (partial responses plus stable disease) was 96% (25 of 26 patients). Figure 3 indicates that PFS did not differ between patients who were selected by histology (9.2 months) or cytology (8.6 months).

Figure 3.

These are Kaplan-Meier curves of progression-free survival (PFS) from gefitinib-treated patients. There was no difference between patients who were selected for treatment by histology (HISTO) and those who were selected for treatment by cytology (CYTO) (P = .88).


This study demonstrates that the detection of EGFR mutations by a centralized laboratory on routine cytologic specimens is as reliable as histologic testing in predicting the response to gefitinib treatment. In fact, among patients who were tested on histology, the RR was 54% and the PFS was 9.2 months. Patients who were selected on cytologic samples had similar outcomes (RR, 62%; PFS, 8.6 months). Consequently, despite concerns that less-than-optimal samples are sent to reference molecular laboratories, cytologic samples can reliably predict treatment outcome. Previous studies demonstrated that, technically, EGFR mutations can be reliably detected on paucicellular lung cancer cytology specimens.[6, 9-12, 20] Our study also demonstrates that paucicellular cytology samples can predict the efficiency of gefitinib treatment.

Externalizing EGFR testing is fraught with a high rate of inadequate samples. In fact, Pang et al reported that 10% of all samples were inadequate.[8] Despite the use of a very sensitive methodology (the TheraScreen EGFR29 kit; Qiagen, Hilden, Germany), 14% of samples received by Allegrini et al were deemed inadequate.[6] Although our approach required the presence of only 25 neoplastic cells, 16% of cytologic samples were rejected by our laboratory because the level of tumor cellularity was below the analytic sensitivity of our procedure. To improve the rate of adequate samples, primary (cyto)pathologists should be engaged as knowledgeable partners in the molecular diagnostic process.[21] The relevance of optimal cytopreparation cannot be overemphasized. Given well prepared samples, cytology has several notable advantages over bioptic specimens for NSCLC subtyping, and immunocytochemistry is required only in few cases; thus, molecular testing can exploit most of the available material.[22, 23]

Our study design was similar to those in the studies by Oshita el al[15] and Lozano et al.[16] The former group reported a 91% RR in the 11 patients who had mutations detected on cytologic samples,[15] whereas Lozano et al reported an RR of 75% and a PFS of 12.3 months in 16 patients.[16] However, our approach differed from those studies in at least 2 aspects. First, our samples came from external institutions; and, second, unlike the studies by Oshita et al and Lozano et al, our 2 groups of patients were tested with the same molecular assay, and all received treatment with gefitinib. The only between-group difference was the type of sample (histology or cytology) tested.

In 2 of our samples, the deletion in exon 19 was detected by the fragment assay only and was missed by direct sequencing (Fig. 2). Indeed, methods of differential sensitivity are not always concordant. In previous studies, EGFR mutations detected by mutant-enriched PCR in effusion samples were not detected by nonenriched assays.[24, 25] Similarly, the sensitive TheraScreen EGFR29 kit detected more mutant cases than direct sequencing.[26] Little is known about the clinical behavior of patients who have discordant results at differential sensitivity testing. Both of our patients who had discordant results had a partial response to gefitinib treatment. Therefore, it is conceivable that gefitinib treatment is effective even when a limited number of mutated cells are detected thanks to the use of mutation detection techniques that are more sensitive than direct sequencing. However, Han et al reported clinical progression in a patient whose EGFR mutation was detected by highly sensitive methods, whereas direct sequencing indicated a WT EGFR status.[14]

In conclusion, prospective studies are needed to determine the correlation between genetic intratumoral heterogeneity, mutational detection tools, and gefitinib treatment outcomes. Our current results provide patients, clinicians, and requesting primary pathologists data demonstrating that treatment prediction based on EGFR mutational diagnosis performed by a centralized laboratory on cytologic material has a degree of reliability similar to that provided by testing histologic samples.


No specific funding was disclosed.


The authors made no disclosures.