Phase II Study of Gefitinib in Patients with Relapsed or Persistent Ovarian or Primary Peritoneal Carcinoma and Evaluation of Epidermal Growth Factor Receptor Mutations and Immunohistochemical Expression: A Gynecologic Oncology Group Study

Purpose: This phase II trial assessed the activity and tolerability of a daily oral dose of 500 mg gefitinib (ZD1839, Iressa) in patients with recurrent or persistent epithelial ovarian or primary peritoneal carcinoma, and explored the clinical value of determining the status of the epidermal growth factor receptor (EGFR).

Experimental Design: Primary measure of efficacy was progression-free survival at 6 months. Mutations in exons 18 to 21 of EGFR and/or immunohistochemical expression of EGFR were evaluated in tumor specimens from patients enrolled in this trial as well as from patients not treated with gefitinib.

Results: Twenty-seven of 30 (90%) patients were eligible and evaluable for analysis of gefitinib efficacy and toxicity. Of these, four survived progression-free >6 months with one objective response (4%). The most commonly observed grade 3 toxicities were dermatologic (15%, 4 of 27) and diarrhea (30%, 8 of 27). Specimens from 26 of 26 or 25 of 26 patients were evaluable for immunohistochemical or mutation analysis, respectively. The response rate for patients with EGFR-positive tumors was 9% (1 of 11). EGFR expression was associated with longer progression-free survival (P = 0.008) and possibly longer survival (P = 0.082). The patient with the only objective response had a mutation in the catalytic domain of the tumor's EGFR (P = 0.04). Among 32 invasive tumors from patients not treated with gefitinib, one exhibited a catalytic domain mutation.

Conclusions: Gefitinib was well tolerated but had minimal activity in unscreened patients with recurrent ovarian or primary peritoneal carcinoma. Prescreening patients for activating mutations in EGFR may improve response rate to gefitinib. This report is the first to document activating mutations in catalytic domain of EGFR in 3.5% (2 of 57) of ovarian cancers.

Epithelial ovarian cancer is the most lethal of gynecologic malignancies in the United States. It is the fourth leading cause of cancer death among women with an estimated 22,220 new cases and 16,200 deaths in 2005 (1). Approximately 75% of patients with this form of ovarian cancer are diagnosed at an advanced stage with the disease having already spread throughout the peritoneal cavity. Whereas there has been an improvement in the 5-year survival rate for patients with stage III disease, the long-term survival rate remains ∼15% to 25% (2). Recurrences are common and subsequent responses to second-line chemotherapy are often of short duration. Therefore, the development of new therapies remains a critical need.

Growth factors and their receptors are known to play critical roles in cell growth and differentiation (3). Many of these receptors possess tyrosine kinase activity that is activated upon the binding of the receptor with its ligand (4). Growth factors, such as epidermal growth factor (EGF), are potent mitogens for several human epithelial cancer cell types, including ovary, and have been implicated in cancer development (5). Expression of two or more EGF-related proteins or EGF receptor (EGFR) is significantly correlated with increased surgical stage in serous and clear cell ovarian carcinomas (6). EGFR is expressed in 35% to 70% of ovarian cancers (7, 8). Specimens obtained postcisplatin treatment compared with specimens from the same patients before chemotherapy exhibited more intense and diffuse expression of EGF-related proteins and EGFR. In addition, EGFR overexpression has been found in human cancer cell lines that are resistant to cytotoxic agents. Thus, interruption of the mitogenic signaling pathway associated with EGFR tyrosine kinase activity is likely to inhibit cell proliferation of malignant tumors (7, 8).

Clinical success with inhibitors of the ERBB family of receptors, such as trastuzumab (Herceptin), in blocking ERBB2 has already been documented (9). This agent potentiates the efficacy of cytotoxic chemotherapy when given concurrently. More recently, cetuximab, a monoclonal antibody against EGFR, was shown to sensitize tumors in patients with relapsed or refractory disease to chemotherapeutic agents to which they were previously resistant (10). This drug recently was approved by the Food and Drug Administration in the treatment of colon cancer in combination with irinotecan.

Gefitinib (ZD1839, Iressa) is a low-molecular weight (447 Da) quinazoline derivative that specifically inhibits the activation of EGFR tyrosine kinase through competitive binding of the ATP-binding domain of the receptor (11). This compound is reported to be ∼100-fold less active against ERB2 tyrosine kinase and has no inhibitory activity against other tyrosine or serine-threonine kinases (12–15). Gefitinib treatment results in reversible inhibition of EGF-stimulated autophosphorylation of EGFR in a variety of EGFR-expressing human cancer cell lines (16). Mechanistically, gefitinib treatment was associated with cell cycle arrest at the G₀-G₁ boundary in a dose- and time-dependent manner that involved the increased expression of the cyclin-dependent kinase inhibitor p27^KIP1 (17). The inhibition of non-EGFR–stimulated cell growth is reported to require a 40-fold higher concentration of gefitinib, confirming the specificity of this compound (18). Similar preclinical activity was shown in vivo by the inhibition of tumor growth in mice bearing HX62 ovarian carcinoma xenografts (18). Rapid regrowth was observed in the majority of mice when the drug was discontinued, suggesting that chronic administration will be required to maintain control of tumor proliferation. Gefitinib treatment was also shown to be associated with similar antitumor activity among xenograft models with variable EGFR expression, suggesting that other factors besides EGFR expression may predict tumor response to gefitinib. These factors could include expression levels of activating ligands, dimerization partners, the presence of mutated receptors, and the activation of downstream signaling pathways by EGFR.

Gefitinib enhances the effects of a variety of chemotherapeutic agents, including platinum compounds, taxanes, topoisomerase I and II inhibitors, and anthracyclines, in various human tumor cell lines from ovary, breast, and colon malignancies that express the EGFR (19, 20). It has been shown to be well tolerated with predictable pharmacokinetic parameters in phase I trials (21–23). Based on the data from these early trials, an oral fixed dose of 500 mg of gefitinib was chosen for evaluation for this trial. After the trial had commenced, subsequent data from trials in patients with lung cancer established 250 mg as the recommended dose (24, 25). Single agent gefitinib was approved by the Food and Drug Administration for third-line treatment of non–small cell lung cancer (NSCLC). Erlotinib, a similar EGFR inhibitor, showed a level of activity in patients with recurrent ovarian carcinoma similar to the results reported in patients with NSCLC (26).

The present phase II trial was undertaken to assess the activity and tolerability of gefitinib in patients with recurrent or persistent ovarian carcinoma or primary peritoneal cancer. In addition, testing was done to evaluate the expression of EGFR by immunohistochemical and mutational status within the tyrosine kinase domain of the receptor in archival tumors from the patients in this trial and to explore the association between these biomarkers and clinical outcome in a manner that was similar to reports in patients with NSCLC treated with gefitinib. These reports in NSCLC showed a strong association between responses to gefitinib and the presence of such mutations in the EGFR (27, 28).

Eligibility. Eligible patients had documented epithelial ovarian or primary peritoneal carcinoma. Confirmation of persistent or recurrent disease was required and could be documented either clinically or histologically. The tumors of these patients were not required to be EGFR positive to be eligible for enrollment onto the trial. Patients were required to have measurable disease; a Gynecologic Oncology Group (GOG) performance status of 0 to 2; and adequate bone marrow (absolute neutrophil count ≥1,500/μL, platelet count ≥100,000/μL), renal (serum creatinine ≤1.5 times the upper limit of normal), and hepatic function (total bilirubin ≤1.5 times upper limit of normal and transaminases and alkaline phosphatase ≤2.5 times upper limit of normal). Eligible patients were permitted to have had up to two prior cytotoxic regimens and were required to have platinum-free interval of <12 months. Patients with active corneal disease, unstable cardiac disease, symptoms consistent with bowel obstruction, prior therapy with an EGFR antagonist, or ingesting CYP3A4-inducing agents (e.g., phenytoin, carbamazepine, barbiturates, nafcillin, or St. John's Wort) were excluded. Patients provided written informed consent consistent with federal, state, and local institutional requirements to participate in the clinical and translational research components of this protocol before receiving any protocol therapy. In addition, the protocol was approved by the institutional review board at each of the participating GOG institutions and done in accordance with assurances filed with and approved by the Department of Health and Human Services. Histologic diagnosis for each patient enrolled on this protocol was confirmed by members of the GOG Pathology Committee.

Treatment plan and dose modifications. The initial dose of gefitinib (AstraZeneca, Wilmington, DE) was a fixed daily dose of 500 mg orally until progressive disease or adverse effects prohibited further therapy with this agent. A cycle equaled 28 days. Toxicity was graded using the National Cancer Institute Common Toxicity Criteria Version 2.0. Patients who experienced grade ≥3 rash or grade ≥2 nonhematologic toxicity had therapy held until resolution to a grade ≤1. Then, gefitinib was restarted at 250 mg daily. There was no dose reduction below 250 mg. Once a patient's dose was reduced, it was not subsequently increased. Treatment delays of >14 days required removal from protocol treatment.

Response assessment. Patients were clinically evaluated every 4 weeks and radiologically every 8 weeks. The same evaluation modality was used throughout for each patient on the study. Response criteria were used as defined by Response Evaluation Criteria in Solid Tumors (29).

Tumor specimens. Patients participating in this gefitinib trial were required to provide both fixed and frozen tumor specimens for laboratory testing. Archival formalin-fixed and paraffin-embedded tumor tissue, a block or unstained sections, from the primary diagnosis was obtained to examine immunohistochemical expression and mutational status of EGFR. Fixed and frozen tumor biopsies were prepared before initiating gefitinib treatment and after completing two cycles (∼8 weeks) of gefitinib. Biopsies were done using a 16-gauge or larger bore needle unless a more accessible tumor allowing an incisional biopsy was available. A portion of the tissue biopsies was instantly frozen in ornithine carbamyl transferase compound or directly in liquid nitrogen and stored at ≤−70°C, and the remaining part was fixed in buffered formalin and embedded in paraffin. The paired pre- and posttreatment biopsies were obtained to evaluate expression of EGFR, phosphorylated EGFR, and downstream signaling pathways. However, because only a few matched pairs of samples were accrued, no meaningful analysis could be done on the pretreatment and posttreatment biopsies.

Additional archival and fresh-frozen tumor specimens were obtained from patients who were not treated with gefitinib. The Fox Chase Cancer Center Tumor Bank and Biosample Repository provided archival formalin-fixed and paraffin-embedded tumor tissue from 130 ovarian cancer patients organized in a tissue microarray for this study to assess immunohistochemical expression of EGFR in a larger spectrum of ovarian cancers, and fresh-frozen tumor specimens from 32 patients with invasive ovarian cancer, 8 patients with a benign ovarian tumor, and 10 patients with borderline ovarian cancer (tumors with low malignant potential) to evaluate EGFR catalytic domain mutations. Informed consent was obtained from each of these Fox Chase Cancer Center patients to collect and use their tumor specimens for research, and approval from the institutional review board at Fox Chase Cancer Center was obtained before any laboratory testing was done.

Mutational analyses of epidermal growth factor receptor in ovarian tumors. Genomic DNA was isolated from the archival formalin-fixed, paraffin-embedded tumor tissues from patients treated with gefitinib or from fresh frozen tumor specimens from patients who were not treated with gefitinib. EGFR exons 18 to 21 were amplified by PCR using primary and secondary PCR primer pairs published previously (27). PCR fragments were cleaned with QIAquick PCR Purification kit (Qiagen, Inc., Valencia, CA) and sequenced and analyzed in both sense and antisense directions for the presence of heterozygous mutations. Analysis of the DNA sequence was done using Sequencher v4.2 (Gene Codes Corp., Ann Arbor, MI). All sequence variants were confirmed by multiple independent PCR amplifications using independent isolates of DNA. Mutational analyses were done without knowledge of clinical outcome including tumor response.

Subcloning the PCR product and sequence analysis. Because the PCR products for EGFR exon 19 (GO43) showed two bands on an agarose gel, each band was purified as described above and was subcloned directly into pCR 4-TOPO vector using TA Cloning kit for sequencing (Invitrogen, Carlsbad, CA) following the instruction of the manufacturer. PCR was done to identify bacterial colonies containing clones with appropriate inserts. Plasmid DNA was purified using QIA filter Plasmid Maxi kit (Qiagen) and sequenced using a universal M13 primer.

Immunohistochemistry assay for epidermal growth factor receptor. An immunohistochemistry assay was done on archival formalin-fixed, paraffin-embedded tissues after deparaffinization through various grades of alcohol solution to assess EGFR expression. The tissue was treated with 0.05% pepsin (Sigma, St. Louis, MO; pH 2.0) at 37°C for 20 minutes for antigen retrieval. The tissue was then incubated with 1:40 dilution of anti-EGFR monoclonal antibody (Zymed, South San Francisco, CA) at room temperature for 30 minutes. Bound primary antibody was detected using Envision Mouse Antibody Detection kit (DAKO, Carpinteria, CA) and ABC techniques following the recommendation of the manufacturer. All slides were processed using a Techmate 1000 or Benchmark XT (Ventana Medical, Tucson, AZ) automated immunostainer. The reviewer (H. Wu), a board-certified pathologist, evaluated each of the slides by light microscopic examination to assess staining intensity (0, 1+, 2+, or 3+) and the approximate percentage of positively stained tumor cells without any knowledge of clinical outcome including tumor response. Absence of any specific membranous staining within the tumor was scored as 0. Positive staining is defined as any staining of the tumor cell membranes above the background level, whether it is complete or incomplete circumferential staining. Staining intensity was further scored as 1+ (low), 2+ (moderate), and 3+ (strong). The fraction of positive tumor cells was estimated over the entire tissue section available for review.

Statistical methods. The trial used a two-stage accrual design with an early stopping rule in the event that the treatment showed insufficient activity (30). The primary end point used to evaluate drug efficacy was progression-free survival at 6 months. The time at risk for progression or death was measured from the date of entry into the clinical trial. Those eligible patients who survived progression-free for at least 6 months were considered treatment successes, whereas those who did not or were censored before 6 months were considered treatment failures. During the first stage of accrual, 22 to 29 patients were to be entered and evaluated. If >3 of 22 to 24 patients or 4 of 25 to 29 patients survived progression-free for 6 months or longer, then a cumulative accrual of 53 to 60 patients would be attained in a second stage. The treatment would be considered worthy of further investigation if >10 of 53, 11 of 54 to 57, or 12 of 58 to 60 patients survived progression-free for 6 months or more.

If the true probability of being alive and progression-free after 6 months is 15% (a condition believed to be associated with ineffective treatment for this population), these decision rules limit the average probability of designating the treatment as active to 10%, and the average probability of stopping after completing only the first stage of accrual is 59%. On the other hand, if the true probability of this type of response is 30% (an indication of activity that is considered clinically significant), then the average probability of correctly classifying the treatment as active is 90%. These statistical quantities are average probabilities computed from individual probabilities over all permitted accrual combinations and assuming each combination is equally likely. Limited investigations have indicated that the type I and type II errors are fairly insensitive to variations in the true probability distribution of accrual combinations (30).

Determination of the nature and degree of toxicities associated with gefitinib was an additional primary objective of this trial. In addition, the secondary end points used to evaluate drug efficacy included overall survival, overall progression-free survival, and frequency of response (partial and complete response). Overall survival and progression-free survival from patients on this study were characterized by quartile estimates, plotted with Kaplan-Meier estimates, and compared with historical controls. Exploratory data analyses were done to evaluate associations between immunohistochemistry results or mutational analyses and clinical outcomes. Patient age, GOG performance status, tumor grade, and platinum resistance were evaluated as potential explanatory variables for the translational research findings. Statistical analysis of these data was done using SAS for Window version 9.1 software (SAS Institute, Inc. Cary, NC).

Patients and eligibility. Thirty patients were enrolled onto the trial. Three patients were not eligible (one had a cancer other than epithelial ovarian or primary peritoneal carcinoma and two had a platinum-free interval >12 months). Patient characteristics are listed in Table 1. Twenty-five patients had a GOG performance status of 0 or 1. All but two eligible patients had papillary serous tumors. Twenty-three patients had an ovarian primary and four patients had a primary peritoneal carcinoma. The number of prior cytotoxic regimens was nearly evenly distributed with 13 receiving one prior regimen and 14 exposed to two prior regimens. Seventeen of 27 (63%) patients were considered to have platinum-resistant disease (<6 months off prior platinum therapy).

Treatment responses. Patients received a median of two cycles (range, 1-37) of protocol therapy. Only four patients had a progression-free survival ≥6 months including one partial responder, thus the study did not continue to second-stage accrual (Table 2). Of these four, two patients had one prior treatment regimen and two had two prior regimens. The median progression-free survival for the whole group was 2.17 months (first quartile, 1.54 months; third quartile, 3.81 months; Fig. 1). The patient who achieved a partial response had a progression-free survival of 26.94 months. This patient is still alive but is now receiving other chemotherapy. One patient with stable disease is still progression-free at 32.26 months. Of note, the patient with a progression-free survival of 9.95 months had only one cycle of study treatment that was discontinued due to grade 3 skin rash. The remaining patient who met the primary end point had a progression-free survival of 9.07 months. The median overall survival was 12.2 months (first quartile, 4.4 months; third quartile, 27.89 months; Fig. 1). There was no relationship between dose of gefitinib and response. Eight of 21 patients who progressed within 6 months had dose reductions (data from the two patients who progressed within 6 months of starting gefitinib are not available). Four of these patients had dose reductions during the first cycle and the other four had a dose reduction during the second cycle. The four patients who had progression-free survival >6 months had a dose reduction at 1, 6, 10, and 20 months.

Toxicity. All 27 eligible patients were included in the toxicity analysis. The most commonly observed grade 3 toxicities were dermatologic and gastrointestinal (Table 3). Hematologic and neurologic toxicities were minimal. Overall, there was little evidence of an association between dermatologic toxicity and surviving progression-free for >6 months. Grade 3 rash was associated with improved outcome (Table 4). Diarrhea (overall or of any specific grade) was not associated with outcome.

Baseline expression of epidermal growth factor receptor in ovarian tumor samples. Archival tumor from the primary diagnosis was submitted for 26 of the 27 eligible and evaluable patients who participated in this trial to assess immunohistochemical expression of EGFR. Positive EGFR expression (+1 or greater) was observed in 11 (42%) of the tumors examined, and 15 (58%) of the cases were deemed negative (as described in the Materials and Methods section). Among the EGFR-positive tumors, 45.5% (5 of 11) displayed low EGFR levels (1+). Of these, staining was observed in <1% tumor cells in three cases, in 20% tumor cells in one case, and 40% of tumor cells in one case. In addition, there were 36.4% (4 of 11) with moderate EGFR levels (2+) in 5%, 10%, 20%, or 40% of tumor cells and 18.2% (2 of 11) with high EGFR levels (3+) in 50% or 90% of tumor cells.

All four patients with prolonged progression-free survival had EGFR-positive tumors although three of the four patients had 1+ or 2+ intensity in <10% of tumor cells, including the responder. Exploratory analyses revealed a statistically significant association between positive EGFR expression and progression-free survival (Fig. 2A). Because the sample size was small, a permutation analysis using the log-rank test estimated the P value to be 0.008. The association between EGFR expression and survival was suggestive (P = 0.082, permutation analysis log-rank test; Fig. 2B). See Table 5 for response data by EGFR expression. There was no evidence of a difference in progression-free survival or survival when patients from this study were compared with historical controls from phase II treatment protocols conducted by the GOG in patients with platinum-free interval <12 months in persistent or recurrent epithelial ovarian or primary peritoneal carcinoma treated with one prior cytotoxic regimen (data not shown). In addition, there was no evidence of an association between platinum resistance status and EGFR expression (data not shown). Patients who expressed EGFR on this study tended to be older. EGFR-negative patients had a mean age of 55, whereas EGFR-positive patients had a mean age of 66. SD in both groups was ∼12 (P = 0.028, ANOVA). The association of EGFR expression and longer progression-free survival or survival could not be explained by this prognostic indicator. All four patients who had prolonged progression-free survival had platinum resistant disease.

Immunohistochemical expression of EGFR was evaluated in an independent panel of 130 primary ovarian tumors organized in a tissue microarray to determine whether the expression profile observed in the patients treated with gefitinib was generalizable beyond this trial. Forty-nine percent (64 of 130) of these tumors exhibited some degree of EGFR expression (1+ or greater) consistent with 42% EGFR-positive cases reported in this phase II trial. The majority of EGFR-positive tumors (38 of 64; 59.4%) showed low levels (1+) of the receptor, whereas 37.5% (24 of 64) were 2+ and only 3.1% (2 of 64) were judged 3+.

Mutational analysis of epidermal growth factor receptor in ovarian tumors. Mutations in the tyrosine kinase domain region of EGFR have been reported in NSCLC that are sensitive to gefitinib (27, 28, 31). We hypothesized that patients with ovarian cancer who respond to gefitinib would have similar somatic mutations in the EGFR gene, indicating the essential role of the EGFR signaling pathway in the tumor. DNA suitable to evaluate somatic mutations in the EGFR gene was able to be extracted from 25 of the 26 archived tumor tissue specimens submitted from the eligible and evaluable women in this trial. To avoid selection bias, the samples were evaluated in a blinded fashion. Based on previous studies (27, 28, 31), we focused our mutation analysis on the tyrosine kinase domain encoded by exons 18 to 21 of EGFR. Overall, one tumor possessed a tyrosine kinase domain mutation among the 25 (4.0%) tumor DNAs examined. As shown in Fig. 3A, one variant band was detected in the responding patient. The lower band was found to contain an in-frame deletion removing 15 nucleotides (2235del15; E746_A750del; Fig. 3B). Matched normal tissue was not available for this patient, but previous studies of NSCLC showed that the 2235del15 mutation had arisen somatically during lung tumor formation (27, 28). Following completion of the mutational studies, the code for the samples was broken, and it was determined that the patient with the mutated EGFR in her tumor experienced a partial response. This patient received 29 cycles of gefitinib and had a progression-free survival time of ∼27 months. By comparison, no mutations were observed in 24 of the cases from which DNA was retrievable and who had no measurable response to gefitinib (P = 0.04, exploratory one-sided Fisher's exact test). However, one patient with stable disease who received 37 cycles who had a long progression-free survival time did not have a detectable mutation in exons 18 to 21 of EGFR.

Given the apparent clustering of EGFR mutations based on previous studies of NSCLC, we sequenced exons 18 to 21 in an additional 32 ovarian adenocarcinomas, 8 benign ovarian tumors, and 10 borderline ovarian cancers from patients who did not participate in this trial. The majority of the adenocarcinomas represented advanced stage disease (21 papillary serous, 4 mucinous, 2 endometrioid, 2 undifferentiated, 2 mixed Müllerian, and 1 clear cell). One additional ovarian cancer sample was found to have a mutation (3.1%; 1 of 32), whereas no mutations were observed in benign or borderline tumors (0 of 18). Interestingly, the mutation detected was in exon 19 and was identical to one of the two mutations found in the patient who responded to gefitinib (i.e., 2235del15; E746_A750del; Fig. 4A and B). Sample 22 was a grade 3, poorly differentiated papillary serous cystadenocarcinoma. The metastatic tumor was ER/PR and HER-2 negative, but positive for p53 tumor suppressor gene product by immunohistochemistry. What was unique compared with NSCLC was that the tumor was homozygous for the mutation. Evaluation of constitutional DNA from this individual showed that this was a somatic mutation (data not shown) and semiquantitative PCR of the tumor DNA did not suggest EGFR gene amplification. Overall, we detected mutations in the tyrosine kinase domain region in 3.5% (2 of 57) of ovarian adenocarcinomas and observed that the patient with the only objective response had a mutation in the catalytic domain of the EGFR of tumor (P = 0.04), suggesting that prescreening patients for activating mutations in EGFR may improve response rate to gefitinib.

Tyrosine kinases regulate signaling pathways that control critical cellular activities. When overexpressed or activated by mutations, tyrosine kinases can contribute to the development of cancers. If tumor cells depend on a mutant tyrosine kinase for survival, then the mutated enzyme can fortuitously serve as an “Achilles' heel” for cancer therapy. Gefitinib was the first agent designed with a known molecular target to receive Food and Drug Administration approval for the treatment of lung cancer targeting the tyrosine kinase receptor, EGFR, yet its activity is limited to a subgroup of patients with NSCLC (32–34). Recently, studies have shown that activating mutations within the tyrosine kinase domain of EGFR are associated with dramatic responses to gefitinib in this subgroup (27, 28, 31). Before these studies were reported, the GOG completed a phase II trial to assess gefitinib in patients with recurrent epithelial ovarian or primary peritoneal carcinoma. Gefitinib, as a single agent, was well tolerated but had minimal activity in patients with recurrent ovarian or primary peritoneal carcinoma unselected by EGFR status. Toxicities were as expected with skin rash and diarrhea being most prevalent. Skin rash of grade 3 or higher was associated with a positive clinical outcome. This observation is consistent with the findings of other investigators although the number of patients in this study is small (35, 36).

To determine the molecular basis of the action of gefitinib in the patient who responded, archival tumor specimens were tested to evaluate expression of EGFR and mutations in the tyrosine kinase domain of this receptor. To date, reports of mutations involving the tyrosine kinase domain of EGFR have been limited to lung cancer. However, a somatic mutation was detected in the one patient with ovarian cancer who responded to gefitinib in this trial. This patient experienced a partial response and long progression-free survival following gefitinib treatment. Moreover, this patient was found to carry an in-frame deletion (2235del15; E746_A750del) in the catalytic domain of EGFR. This mutation has been reported previously and is one of the more common mutations observed in NSCLC (27, 28, 37). Prolonged stable disease in the other three patients simply may reflect the biology of these three tumors rather than be a consequence of therapy. The former is suggested by the fact that there was no difference in progression-free survival or OS of patients on this trial when compared with patients on other similar GOG trials. A patient with stable disease also experienced a very long progression-free survival, but a mutation in exons 18 to 21 of EGFR was not found in this case. This was not unexpected given that previous studies of patients with stable lung disease also lacked mutations in EGFR (27, 28).⁸

Current efforts are focused on determining whether heterodimerization of wild-type EGFR with other members of the ERBB family may be important in the attenuated responses observed in a subset of patients with stable disease. In support of these studies, Amann et al. (37) have found that beyond the contribution of EGFR mutations, high expression of ERBB family members through gene amplification or other mechanisms with dysregulation of AKT signaling may also contribute to gefitinib sensitivity. Interestingly, gefitinib has been reported to inhibit autophosphorylation of HER2 in breast cancer cells in which HER2 is preferentially activated by heterodimerization with EGFR at near clinically relevant doses (13, 38). Although we detected a possible relationship between progression-free survival and EGFR expression and that all four patients with prolonged progression-free survival had EGFR-positive tumors, there was no evidence to suggest a significant relationship between the approximate percentage tumor cells expressing detectable EGFR expression and progression-free survival (or survival), suggesting that other members of the ERBB family may play a role in clinical outcome. This observation is in contradistinction to recently published data that patients with EGFR-negative colon cancers, as determined by immunohistochemistry, still can respond to cetuximab-based therapy and that testing tumor for EGFR status by immunohistochemistry seems unwarranted (39). This apparent difference between EGFR-positive and EGFR-negative tumor responses in colon cancer for cetuximab and ovarian cancers for gefitinib may be explained by a differential distribution of two distinct EGFR populations. The majority of EGFR are of low affinity (∼95%) and do not contribute significantly to signal transduction (40). A small number of high-affinity EGFR are responsible for the biological activity. It may be that patients with EGFR-negative colon tumors that still respond to cetuximab may have a low number of high-affinity receptors as determined by immunohistochemistry but a sufficient number to be responsive to an EGFR-targeted agent. Ovarian tumors negative for EGFR by immunohistochemistry may be truly negative or below a critical threshold. Additional analyses were conducted to test for associations between EGFR expression and known prognostic factors, such as performance status, platinum sensitivity, and tumor grade. These variables seemed to be independent of EGFR expression. Studies are under way to evaluate these samples in greater detail for evidence of altered expression of other ERBB family members. Of course, such analyses with small sample sizes are fraught with hidden biases and maybe reflective of hidden imbalances in unknown prognostic factors that are beyond the scope of this trial to detect. These results, together with the finding of EGFR mutations in a tumor from a patient with ovarian cancer who had not received gefitinib, suggest that such mutations, although rare (3-4%) in ovarian cancer, may predict a favorable response to gefitinib-based therapy or other tyrosine kinase inhibitors (37). Further study would be required to validate this hypothesis. The long-term fate of gefitinib, however, is uncertain as recently released data from a phase III trial comparing gefitinib to supportive care in patients with recurrent NSCLC showed no survival advantage. The continued approval of gefitinib is currently being reviewed by the Food and Drug Administration. These findings still are relevant to other small molecule tyrosine kinase inhibitors of EGFR that have increased activity in patients with such mutations (31).

Of interest, one of the sporadic ovarian tumors (no. 22) possessing mutant EGFR apparently lacked a wild-type allele, whereas the patient who responded to gefitinib seemed to be enriched for the mutant allele. This is in stark contrast to previous studies of NSCLC (27, 28, 31), which have found both wild-type and mutant alleles in the same tumor. The heterozygous nature of EGFR mutations in these NSCLC tumors is thought to suggest that they exert a dominant oncogenic effect, despite the presence of the second wild-type allele. However, recent studies of NSCLC have found EGFR mutations associated with high levels of EGFR gene amplification in NSCLC cell lines (37). The significance of biallelic activation of EGFR in at least one of the ovarian tumors is currently unknown but suggests a selection for the dominant activated form and that the tumor may be more clinically responsive to gefitinib.

EGFR mutations in the catalytic domain have not been described in ovarian cancers previously; however, we have observed a tumor-specific splice variant, called EGFRvIII, a naturally occurring splice variant that deletes a substantial portion of the subdomain I and II that results in constitutive activation of the receptor (41, 42). Of the seven exons that code for the tyrosine kinase domain (exons 18-24), mutations seem to be limited to the first four exons (exons 18-21; refs. 27, 28, 31, 37, 43), which codes for the smaller N lobe and part of the C lobe, including the crucial activating loop. The mutations identified in this study have been observed in NSCLC and form a tandem around the crucial αC helix in the N lobe (44). These mutations are hypothesized to result in similar configurational changes, causing a shift of the helical axis narrowing the ATP-binding cleft and increasing tyrosine kinase inhibitor sensitivity (27, 28, 44). Although the necessity of ligand binding remains, at least two of the common mutations identified in EGFR (deletions in exon 19 as we have also observed and the L858R missense mutation found in NSCLC) are known to increase the amount and duration of ligand-dependent activation, and explain the much greater sensitivity of mutant cells to gefitinib (27, 45). Furthermore, a recent report by Kobayashi et al. (46) showed that drug sensitivity of NSCLC to gefitinib could be further altered following the acquisition of a second somatic mutation within the kinase region of the EGFR. This latter mutation, within exon 20, lead to the development of gefitinib resistance after the patient with a lung tumor containing an exon 19 mutation had been responding for 24 months. These findings highlight the importance of incorporating periodic biopsies into clinical studies of novel targeted therapies.

The finding of EGFR mutations in ovarian cancer represents a potentially important genetic abnormality because of the promise of immediate clinical application. Although somewhat rare in ovarian cancer, the application of mutational data for clinical care could have a significant impact on a small subset of women with ovarian cancer. In addition, these studies suggest that in response to treatment with gefitinib, ovarian tumors with EGFR mutations will undergo tumor shrinkage, whereas another subset of tumors without EGFR mutations may respond or exhibit stability of disease. Being able to identify these subsets of women with ovarian cancer that are likely to achieve clinically meaningful responses to tyrosine kinase inhibitor, such as gefitinib, before initiating therapy may soon become clinical reality.

The following Gynecologic Oncology Group member institutions participated in this study: Abington Memorial Hospital, Colorado Gynecologic Oncology Group PC, Milton S. Hershey Medical Center, Tufts-New England Medical Center, Rush-Presbyterian-St. Luke's Medical Center, State University of New York Downstate Medical Center, State University of New York at Stony Brook, Columbus Cancer Council, Fox Chase Cancer Center, University of Oklahoma, University of Chicago, Tacoma General Hospital, Thomas Jefferson University Hospital, Case Western Reserve University, and Tampa Bay Cancer Consortium.

We thank Dr. Daphne Bell (Massachusetts General Hospital) for helpful advice; Drs. Michael Bookman and Robert Ozols for their thoughtful comments; Brock Armstrong and Carolyn Slater (Fox Chase Cancer Center) for their technical expertise in the mutational and expression studies; Sandra Dascomb for management of clinical information and help at every stage of this study; Rose Adams-McDonnell for the immunohistochemical analyses of EGFR in tumor samples in the Immunohistochemical Laboratory of the Department of Pathology at Fox Chase Cancer Center under the direction of Dr. Harry S. Cooper; and the DNA sequencing, Tumor Bank, and the Biosample Repository (http://www.fccc.edu/clinicalresearch/BRCF/) core facilities at Fox Chase and the GOG Tissue Bank for support with services and specimens.