Volumetric tumor growth in advanced non-small cell lung cancer patients with EGFR mutations during EGFR-tyrosine kinase inhibitor therapy
Developing criteria to continue therapy beyond RECIST progression
Mizuki Nishino MD,
Department of Radiology, Dana-Farber Cancer Institute and Brigham and Women's Hospital, Boston, Massachusetts
Corresponding author: Mizuki Nishino, MD, Department of Radiology, Dana-Farber Cancer Institute and Brigham and Women's Hospital, 450 Brookline Avenue, Boston, MA 02215; Fax: (617) 582-8574; firstname.lastname@example.org
The objective of this study was to define the volumetric tumor growth rate in patients who had advanced nonsmall cell lung cancer (NSCLC) with sensitizing epidermal growth factor receptor (EGFR) mutations and had initially received treatment with EGFR-tyrosine kinase inhibitor (TKI) therapy beyond progression.
The study included 58 patients with advanced NSCLC who had sensitizing EGFR mutations treated with first-line gefitinib or erlotinib, had baseline computed tomography (CT) scans available that revealed a measurable lung lesion, had at least 2 follow-up CT scans during TKI therapy, and had experienced volumetric tumor growth. The tumor volume (in mm3) of the dominant lung lesion was measured on baseline and follow-up CT scans during therapy. In total, 405 volume measurements were analyzed in a linear mixed-effects model, fitting time as a random effect, to define the growth rate of the logarithm of tumor volume (logeV).
A linear mixed-effects model was fitted to predict the growth of logeV, adjusting for time in months from baseline. LogeV was estimated as a function of time in months among patients whose tumors started growing after the nadir: logeV = 0.12*time + 7.68. In this formula, the regression coefficient for time, 0.12/month, represents the growth rate of logeV (standard error, 0.015/month; P < .001). When adjusted for baseline volume, logeV0, the growth rate was also 0.12/month (standard error, 0.015/month; P < .001; logeV = 0.12*months + 0.72 logeV0 + 0.61).
The characterization of genomic abnormalities in lung cancers in the past decade has transformed the way oncologists approach and treat patients with lung cancer. This is demonstrated best by the discovery and clinical application of epidermal growth factor receptor (EGFR) mutation testing in nonsmall cell lung cancer (NSCLC), which is associated with a dramatic radiographic response to the EGFR tyrosine kinase inhibitors (TKIs) gefitinib, erlotinib, and afatinib.[1-3] Patients with NSCLC who harbor EGFR sensitizing mutations have response rates >70%, and their progression-free survival ranges from 9.7 to 13.1 months when they receive EGFR-TKIs.[4-10] However, virtually all patients who have an initial response eventually progress because of the development of acquired resistance to EGFR-TKIs, and patients demonstrate radiographic tumor growth during TKI therapy.[11-15]
Oncologists have used linear measurements defined according to Response Evaluation Criteria in Solid Tumors (RECIST) as a guide to define response and progression and to determine when to switch therapy or add another agent.[16-18] However, recent clinical observations have indicated that the conventional RECIST-based assessment alone does not fully characterize response and progression in genomically characterized patients who have specific tumor types, including gastrointestinal stromal tumors (GISTs), melanoma, and lung cancers treated with targeted therapies.[19-22] New radiographic response criteria have been proposed, such as Choi criteria for GISTs using computed tomography (CT) density and immune-related response criteria for melanoma, in which a response may be observed after an initial increase in tumor burden.[19-22]
Thoracic oncologists continue to allow patients who have NSCLC harboring EGFR mutations to receive EGFR-TKIs beyond RECIST progression, because their tumors tend to grow slowly, and the patients remain asymptomatic, suggesting that some tumor cells remain sensitive to TKIs.[23-25] Nishie et al demonstrated that continuous EGFR-TKI after progression was associated with improved overall survival compared with switching to chemotherapy alone (hazard ratio, 0.42). EGFR-TKIs are associated with improved quality of life and less toxicity compared with chemotherapy.[7-10] The benefit of EGFR-TKIs should be maximized by adequately prolonging the duration of TKI therapy. A previous study by our group examined patients with NSCLC harboring EGFR mutations who had received a first-line TKI in which 88% of patients continued on TKI beyond RECIST progression, indicating that RECIST progression is not the single determining factor for terminating TKI. There is a clear need for additional radiographic criteria of tumor growth beyond RECIST progression to better guide therapeutic decisions.
One of the major limitations of RECIST is the use of the cutoff value of tumor size increase to define progression, which does not incorporate the changes in tumor burden over time or the tumor growth rate. Tumor volume measurement using multidetector-row CT has been studied to complement the limitations of RECIST. Tumor volume measurements in NSCLC are feasible with higher reproducibility than size measurements.[27-31] We previously established a CT tumor volumetry technique in advanced NSCLC using US Food and Drug Administration-approved software. The study demonstrated that tumor volume was more reproducible than size, consistent with other studies.[27-31] Volume assessment also has been used to predict outcome in patients with NSCLC who received chemotherapy and chest radiotherapy.[32-34] In patients with advanced NSCLC who had sensitizing EGFR mutations, a tumor volume decrease at 8 weeks of EGFR-TKI therapy was associated with longer survival. However, a detailed characterization of volumetric tumor growth rate in patients with EGFR-mutant NSCLC after initial response to EGFR-TKI therapy has not been systematically performed.
Tumor growth is based on a specific relation between tumor volume and time, and comprehensive equations of tumor growth have been pursued extensively in past decades. One of the well studied models is the Gompertzian model, which was described initially by Gompertz in 1825 to deal with human mortality and, unexpectedly, was identified later as useful for describing biologic tumor growth.[36-38] Tumor growth according to the Gompertzian model has an exponential nature at the early stage and subsequently saturates, approaching a plateau as tumor increases. Although the growth of most untreated tumors has been well described by the Gompertzian equation, the growth of treated tumors presents another investigational challenge.[37-39] In the late 1970s, Looney et al quantitatively evaluated tumor growth curves in rat hepatoma during radiotherapy and chemotherapy, attempting to more precisely evaluate therapeutic effects, improve therapeutic scheduling, and better understand tumor biology.[40-42] Those studies, performed more than 3 decades ago, although technologically different, share similar concepts with the current study, in that they focused on the tumor growth rate during therapy to improve response assessment and therapeutic decisions. The objective of the current study was to analyze the volumetric tumor growth rate in patients with advanced NSCLC who had sensitizing EGFR mutations after they reached their volume nadir during EGFR-TKI therapy as an initial step in developing radiographic criteria for slow progression to aid in therapeutic decision making.
MATERIALS AND METHODS
The study population included 58 patients who had stage IV NSCLC (according to the seventh edition of the American Joint Committee on Cancer Cancer Staging Manual) or stage I through IIIA NSCLC with systemic relapse and sensitizing EGFR mutations and who received gefitinib or erlotinib as initial systemic therapy for advanced NSCLC between February 2002 and May 2011 at the Dana-Farber Cancer Institute. The patients had baseline CT scans that demonstrated at least 1 measurable lung lesion (≥10 mm in greatest diameter), at least 2 follow-up CT scans during EGFR-TKI therapy, and experienced volumetric tumor growth during TKI therapy. Patients provided informed consent, and their records were retrospectively reviewed with institutional review board approval.
Tumor specimens were obtained from diagnostic or surgical procedures. Samples consisted of frozen tumor specimens or paraffin-embedded material. EGFR exons 18 through 21 were amplified by polymerase chain reaction and were analyzed bidirectionally by direct sequencing for the presence of somatic mutations.[1, 2] The following EGFR mutations were considered sensitizing: deletions, duplications, and deletion-insertions of exon19, L858R point mutation, L861Q point mutation, and G719 missense point mutations.[25, 43, 44]
Tumor Volume Measurements
Baseline and follow-up CT scans were performed to determine response to EGFR-TKI using the clinical chest CT protocol. Follow-up CT scans were performed every 8 weeks for 33 patients who were enrolled on prospective trials of EGFR-TKIs and at the discretion of the treating providers for 25 patients who received treatment off protocol.[5, 25, 45-47] A thoracic radiologist (M.N.) measured the volume of a dominant, measurable lung lesion (1 lesion per patient) on baseline and all CT scans during EGFR-TKI monotherapy using previously validated, US Food and Drug Administration-approved volume analysis software (Vitrea 2; Vital Images, Minnetonka, Minn). We used this technique based on our previously published data indicating its high interobserver reproducibility, in which tumor volume measurements were more reproducible than size measurements. The nadir (the smallest tumor volume recorded from baseline to TKI termination/last follow-up) was determined for each patient.
In total, 405 volume measurements from the nadir to the end of TKI therapy/last follow-up, with data closure on June 1, 2012, were analyzed. Demographics and disease characteristics were summarized with descriptive statistics. A linear mixed-effects model, fitting time as a random effect, was fitted to the repeated measures of volume data to estimate the effect of time and other prognostic factors on tumor growth. The tumor volume (mm3) was transformed to the natural logarithm scale, and the logarithm of tumor volume (logeV) was used. The first model was built adjusting only for time in months from baseline. Because the baseline volume (logeV0; the tumor volume measured on the baseline scan performed before the initiation of TKI therapy) may influence the tumor volume and its growth rate, the second model was adjusted for time and logeV0. The third model was adjusted for time, for logeV0, and for the clinical characteristics listed in Table 1 to determine whether clinical variables had a significant effect on tumor growth.
Table 1. Patient and Disease Characteristics
No. of Patients (%)
Abbreviations: EGFR, epidermal growth factor inhibitor; del, deletion; NSCLC NOS, nonsmall cell lung cancer not otherwise specified; TKI, tyrosine kinase inhibitor.
Never indicates <100 lifetime cigarettes; former, quit smoking ≥1 year before starting therapy; current: smoked for ≤1 year before starting therapy.
Table 1 summarizes the patient and disease characteristics. The median time on TKI monotherapy was 15.8 months. The median number of follow-up scans was 7.5 (range, 2-35 scans). The median time from baseline to tumor volume nadir was 6.3 months.
Volumetric Tumor Growth Rate
Figure 1 illustrates the volumetric tumor growth of 58 patients from their nadir to the termination of therapy or the last follow-up scan. A linear mixed-effects model was fitted to predict the growth of logeV, adjusting for time from baseline.
In the first model, logeV was estimated as a function of the time from baseline, and the following equation was obtained: logeV = 0.12*time + 7.68.
In this equation, time represents the number of months from baseline. The regression coefficient for time, 0.12/month, represents the growth rate of logeV (standard error, 0.015, P < 0.001).
The second model was adjusted for logeV0 as a fixed effect, and logeV was estimated as follows: logeV = 0.12*time + 0.72 logeV0 + 0.61.
LogeV0 was a significant predictor of the volume after nadir (P < .001), with a coefficient of 0.72. The growth rate of logeV, obtained as a coefficient for time, was 0.12/month (standard error, 0.015, P < 0.001) after adjusting for logeV0. Therefore, the growth rate was 0.12/month for logeV, regardless of the baseline volume.
In the third model, which was adjusted for the clinical variables and for logeV0, stage at diagnosis (stage IV vs others), TKI (gefitinib or erlotinib), and smoking status (current/former vs never smoker) were significant predictors of the volume after nadir, along with logeV0 (P < .001 for logeV0; P = .08 for stage; P = .04 for TKI; and P = .04 for smoking). The following equation was obtained: logeV = 0.12*time + 0.64 logeV0 + 0.83; *stage + 1.00* TKI + 0.54*smoking + 0.34.
In this equation, stage was scored as 1 for stage IV and 0 for stages I through III, TKI was scored as 1 for gefitinib and 0 for erlotinib, and smoking status was scored as 1 for current/former smokers and 0 for never smokers. The growth rate of logeV again was 0.12/month (standard error, 0.01, P < 0.001). Stage, TKI, and smoking status affected the extent of tumor volume; however, the tumor growth rate was 0.12/month for logeV regardless of these clinical characteristics or the baseline volume.
Threshold for Volumetric Tumor Growth
To explore criteria for tumor growth that were appropriate for identifying which patients could safely continue to receive an EGFR-TKI, the threshold of the growth rate of logeV >0.15/month was proposed based on the rate obtained from the equations (0.12/month) plus twice the standard error, 0.015/month (0.12 + 0.015 × 2=0.15), representing the upper 95% confidence interval for the rate. We calculated the growth rate of logeV between 2 consecutive scans after nadir and investigated 2 consecutive occurrences of the growth rate of logeV >0.15/month during EGFR-TKI therapy in all 58 patients.
Fourteen of 58 patients (24%) experienced a growth rate of logeV >0.15/month on 2 consecutive scans, which occurred after nadir in all patients (Figs. 2,3). The median time from baseline to the second scan with a rate >0.15/month was 9.7 months (range, 3.1-20.7 months). The median time on TKI in these 14 patients was 11.7 months, compared with 17.9 months in 44 patients who did not experience a rate >0.15/month on 2 consecutive scans. In 6 of 14 patients (43%), TKI monotherapy was terminated within 1 month from the second scan, and there were no further CT scans before therapy termination. The other 8 patients (57%) remained on TKI monotherapy beyond the second scan and had at least 1 additional chest CT scan while receiving TKI.
The current study provided the volumetric tumor growth rate after nadir in patients with EGFR-mutant, advanced NSCLC who were receiving TKI therapy. The result provides a reference value for the tumor growth rate in patients who progress on TKIs. This growth rate can be studied further in additional cohorts and may help in the development of practical criteria with which to identify patients who can safely remain on EGFR-TKIs. To our knowledge, this is the first report providing a reference value of volumetric tumor growth in a genomically defined cohort of patients with advanced NSCLC who received targeted therapy.
The conventional RECIST-based assessment has limitations in characterizing tumor response and guiding therapeutic decisions in genomically selected cohorts of patients receiving targeted therapy.[19-24] One of the major limitations of RECIST is that it does not take into account tumor growth dynamics or the tumor growth rate, which can be an important factor for characterizing the anticancer activity of targeting agents.
Characterizing tumor growth rates has been a challenging topic for cancer researchers. Since the 1970s, various studies have attempted to characterize tumor growth dynamics of untreated and treated tumors to better understand solid tumor biology and improve therapeutic management.[37-42] Recently, the concept of tumor growth rate during anticancer treatment was studied in a trial of patients with solid tumors to define the optimal trial endpoint.[49, 50] Gomez-Roca et al studied 76 patients with solid tumors who were treated on phase 1 trials, including patients with NSCLC (n = 21), and used tumor volume estimated from tumor size. In that study, the tumor growth rate, obtained as log10(Vt/V0)/dt, decreased by 40% during treatment compared with the pretreatment rate, suggesting that integration of the tumor growth rate may improve the assessment of treatment efficacy. Others studied patients with renal cell carcinoma or prostate cancer and demonstrated that the growth rate constant, obtained as loge2/doubling time using tumor size, was negatively correlated with overall survival.[51, 52]
Those prior studies support the finding that the tumor growth rate adds potentially useful information for assessing response and predicting outcome. The current study focused on patients with EGFR-mutant NSCLC who were receiving TKI, because we believe that the tumor growth rate and its threshold for fast growth versus slow growth are cohort-specific and therapy-specific. We used CT tumor volume measurements rather than tumor size. We chose this approach because we have demonstrated that tumor volume measurement is more reproducible than size measurement and detects smaller changes more accurately. In addition, our group's previous study demonstrated that a tumor volume decrease at 8 weeks of EGFR-TKI therapy was associated with longer survival in patients with EGFR-mutant, advanced NSCLC; whereas tumor size was not associated with survival. To our knowledge, such an approach in this highly specific, genomically defined cohort has never been performed to address the issue of tumor growth rate.
The initial model adjusting for time from baseline alone provided a growth rate of 0.12/month. The growth rate adjusting for time and baseline volume also was 0.12/month, indicating that the baseline volume does influence tumor size at each time point after nadir, although it does not have much effect on the pace of tumor growth; the tumors in our cohort grew at an overall rate of 0.12/month for logeV after their nadir, regardless of their baseline volume. This result is consistent with our prior observation that baseline volume was not associated with survival in patients with EGFR-mutant NSCLC who were receiving EGFR-TKIs. The growth rate also was 0.12/month when the analysis was adjusted for clinical variables, demonstrating the stability of the model. The model and the consistency of the growth rate of 0.12/month need to be validated in a larger, independent cohort of patients with sensitizing EGFR mutations who received TKIs to determine whether the consistent results either are caused by an artifact of the model fitting or truly reflect the biologically driven behavior of EGFR-mutated tumor.
Our goal was to identify a cutoff value capable of differentiating patients who are slowly progressing and can safely remain on EGFR-TKI therapy. The upper limit of the 95% confidence interval for the rate, 0.15/month, was used because it is often used to determine outliers in tumor volume studies and in growth models. We investigated the occurrence of 2 consecutive observations of a rate >0.15/month; only 1 observation may be the result of measurement variability rather than true tumor change, and clinicians tend to give the “benefit of the doubt” and confirm observations on 1 more scan(s) before making decisions. The concept of confirmation is well established in RECIST and is given more emphasis in immune-related response criteria, in which confirmation is required for progression.[16-19]
Two consecutive events of a growth rate >0.15/month occurred after the volume nadir in all 14 patients, which is reassuring. The time on TKI was shorter in these 14 patients than in the 44 patients who did not have such events, indicating that the events did not happen by chance. Information about the tumor growth rate was not available for the providers who treated these patients. It is necessary to validate the model in an independent cohort to apply the threshold prospectively.
Limitations of the current study include a retrospective design and a small number of patients treated at a single institution. Currently, we are planning to expand the cohort and validate our results in an independent cohort to establish practical criteria. The tumor volume was measured in 1 dominant lung lesion per patient, and smaller lung lesions or extrapulmonary lesions were not included, which is another limitation. We designed the study in this way because we believe tumor volume analysis should be additive to the evaluation of systemic tumor burden by RECIST-based approach and by clinical assessment.[20, 35] An ongoing, multicenter, phase 2 trial (National Clinical Trials identifier NCT01310036; the ASPIRATION study) allows continuation of erlotinib beyond RECIST-progressive disease (PD) at the investigator's discretion. Scenarios in which erlotinib may be continued include PD after >6 months of partial response/stable disease, asymptomatic minimal PD, or new brain metastasis controlled locally. Scenarios in which erlotinib should be discontinued are symptomatic extracranial PD, rapid PD and/or worsening of performance status, or life-threatening complications. The objective of our current analysis was to provide a quantitative reference value that can be used along with these clinical criteria to better aid therapeutic decisions and maximize the benefits of effective targeted therapy.
In conclusion, the tumor volume analysis was able to define volumetric tumor growth after the nadir in patients with EGFR-mutant, advanced NSCLC who were receiving EGFR-TKI. The study provided a reference value for the tumor growth rate in patients who progress after the nadir during TKI therapy. Further investigation is warranted to validate these results and to develop practical radiographic criteria that can help identify patients as slow progressors who can safely remain on EGFR-TKIs.
The investigators were supported by grant 1K23CA157631 from the National Cancer Institute (NCI) (M.N.), by grants 1RO1CA114465-07 (B.E.J./P.A.J.) and 5R21 CA11627-02 (H.H.) from the National Institutes of Health (NIH), by grant 2P50CA090578-10 (B.E.J./P.A.J.) from the NCI Specialized Program of Research Excellence (SPORE) in Lung Cancer, by a grant from Genentech Inc., by a grant from the Doris and William Krupp Research Fund in Thoracic Oncology, and by an American Society of Clinical Oncology Translational Research Professorship.
CONFLICT OF INTEREST DISCLOSURES
Dr. Nishino has received support from an Eleanor Shore Fellowship and a Radiological Society of North America Research Scholar grant. Dr. Dahlberg has received salary support from Dr. Nishino's NIH grant. Dr. Hatabu has received grants from Toshiba Medical Inc. and AZE Inc. Dr. Jackman has received compensation as a consultant to Genentech, Foundation Medicine, Inc., and Infinity Pharmaceuticals and has received fees for lectures from Chugai Pharma. Dr. Jänne has received compensation as a consultant and for drug development from Boehringer Ingelheim, Roche, Genentech, Abbott, AstraZenaca, Pfizer, and Sanofi and receives postmarketing royalties from LabCorp, a Dana-Farber Cancer Institute-owned intellectual property concerning EGFR mutations. Dr. Johnson has received research project (RO1) and SPORE grants from the NIH; has received compensation as a consultant from AstraZeneca and Genentech; serves on the advisory boards of Genentech and AstraZeneca; has received payment for a patent on EGFR mutation testing as an indication for EGFR-TKI therapy; and receives postmarketing royalties for EGFR mutation testing from Dana-Farber Cancer Institute for the licensed technology.