Mixed Responses to Systemic Therapy Revealed Potential Genetic Heterogeneity and Poor Survival in Patients with Non‐Small Cell Lung Cancer

Abstract

Background

A subset of patients with non‐small cell lung cancer (NSCLC) fosters mixed responses (MRs) to epidermal growth factor receptor (EGFR)‐tyrosine kinase inhibitors (TKIs) or chemotherapy. However, little is known about the clinical and molecular features or the prognostic significance and potential mechanisms.

Methods

The records of 246 consecutive patients with NSCLC receiving single‐line chemotherapy or TKI treatment and who were assessed by baseline and interim positron emission tomography/computed tomography scans were collected retrospectively. The clinicopathological correlations of the MR were analyzed, and a multivariate analysis was performed to explore the prognostic significance of MR.

Results

The overall incidence of MR to systemic therapy was 21.5% (53/246) and predominated in patients with stage IIIB–IV, EGFR mutations and those who received TKI therapy (p < .05). Subgroup analyses based on MR classification (efficacious versus inefficacious) showed significant differences in subsequent treatment between the two groups (p < .001) and preferable progression‐free survival (PFS) and overall survival (OS) in the efficacious MR group. Multivariate analyses demonstrated that the presence of MR was an independent unfavorable prognostic factor for PFS (hazard ratio [HR], 1.474; 95% confidence interval [CI], 1.018–2.134; p = .040) and OS (HR, 1.849; 95% CI, 1.190–2.871; p = .006) in patients with NSCLC. Induced by former systemic therapy, there were more T790M (18%), concomitant EGFR mutations (15%), and changes to EGFR wild type (19%) in the MR group among patients with EGFR mutations, which indicated higher incidence of genetic heterogeneity.

Conclusion

MR was not a rare event in patients with NSCLC and tended to occur in those with advanced lung adenocarcinoma treated with a TKI. MR may result from genetic heterogeneity and is an unfavorable prognostic factor for survival. Further studies are imperative to explore subsequent treatment strategies.

Introduction

Tailored therapies for patients with non‐small cell lung cancer (NSCLC) are emerging with the prerequisite of comprehensive pathological and molecular testing of potential actionable gene mutations in the new era of personalized medicine. Several randomized studies have consolidated the role of epidermal growth factor receptor (EGFR)‐tyrosine kinase inhibitors (TKI) for patients with advanced EGFR‐mutant NSCLC as a first‐line therapy . Platinum‐based doublet chemotherapies remain optimal for patients with advanced NSCLC with pan‐negative alterations of actionable genes. However, 20%–40% of patients show primarily poor responses to targeted therapies or even chemotherapy. Moreover, most patients eventually develop acquired resistance to drugs, particularly those with EGFR mutations and those taking targeted TKIs. Although second‐line therapies after TKI regimens have been established, especially based on dynamic changes of gene profiles, most resistant mechanisms are suspended for further studies .

Once the disease progresses, not all patients must change to another systemic therapy. Continuing their former regimen combined with local treatment can provide a durable response if well controlled. It is appealing to use genetic or clinical predictors to differentiate patients with a potentially continuous response. Although the origin of resistance is unclear, clinical features of patients may provide valuable hints for further exploration. Almost 38% of patients foster heterogeneous responses to systemic regimens, with the coexistence of responsive and progressive lesions [5]. The organ with mixed response may act as a predictor for second TKIs, with the highest response rate of 77.8% in the central nervous system (CNS) and the lowest in the liver (17.6%, p < .001). This paradoxical phenomenon may be due to intertumoral heterogeneity, pathology, or genetics of primary lung tumors and associated metastases . Tumoral heterogeneity can be induced by targeted therapy or chemotherapy [8] or can originate from intratumoral genetic heterogeneity (ITH) [9], which fosters tumor progression and impedes molecular‐guided treatment. However, ITH of driver genes or significant passenger genes responsible for MR remains controversial and not sufficient enough to guide clinical strategies .

We assessed the occurrence and prognostic significance of MRs to systemic therapies among patients with NSCLC, evaluated by positron emission tomography/computed tomography scans (PET/CT) to provide comprehensive measurements of each lesion. Clinical biomarkers were explored to identify potential molecular mechanisms resulting in mixed response (MR). These results may help us to understand the clinical and molecular features of MR and provide a foundation to explore subsequent therapeutic strategies.

Materials and Methods

Patients and Samples

Patients referred to Guangdong General Hospital, Guangzhou, China, and Guangdong Lung Cancer Institute, Guangzhou, China, and diagnosed with primary pulmonary tumors were reviewed from January 2009 to May 2015. Written informed consent was obtained from all patients for molecular and pathologic analyses of tumor samples in our hospital. This study was approved by the Ethics and Scientific Committee of Guangdong General Hospital. Clinical data were collected from the medical record system. Among 871 evaluable patients, 593 received local treatments, and 32 patients diagnosed with small‐cell lung cancer were excluded. A total of 246 patients with NSCLC evaluated for their response to single‐line chemotherapy or EGFR‐TKIs by PET/CT were enrolled for further analysis. If one patient treated with a certain line of systemic therapy was evaluated by consecutive PET/CT in the time of baseline and progression within this line of treatment, this interval was enrolled. Then, during the subsequent line of systemic therapy after disease progressed, another interval for this patient was also eligible if PET/CT scans were performed at the baseline and disease progression. So, as a result, a patient might be counted twice if consecutive PET/CT scans were available for two lines of therapy. In this setting, 25 patients with two evaluable periods were assessed twice during the administration procedure.

We classified responses to treatment into three parts: responses of primary lesion, responses of metastatic lesions, and presence of new lesions. Based on the corresponding responses to treatment, we defined three types of MRs: (a) MR between primary and metastatic lesions (primary lesions enlarge and metastatic lesions shrink or the opposite; 43.4%), (b) MR between primarily targeted lesions and new lesions (primarily targeted lesions shrink but new lesions present, 34.0%); and (c) MR between separated metastatic lesions (one metastatic lesion enlarges and another shrinks; 22.6%) in the situation that primary lesion was removed. It should be noted that we evaluated the lesions enlarging (+20%) or shrinking (−30%) on the basis of RECIST (equal to partial response [PR] and progressive disease [PD]) (supplemental online Fig. 1). Additionally, there was a low rate of SCLC enrolled in our study. SCLC patients were mostly evaluated first with cranial magnetic resonance imaging (MRI) or electroconvulsive therapy (ECT) scans for high risks of CNS and bone metastases. Then identifications of disease progression by MRI or ECT are quite enough to switch to second‐line chemotherapy. PET/CT scan for those patients are not that common, resulting in fewer patients with consecutive PET/CT scans.

Statistical Analysis

The chi‐square test or the Fisher’s exact test was used to analyze the general clinical characteristics between the two different populations using SPSS 16.0 software (SPSS Inc., Chicago, IL, USA). PFS was measured from the first day of starting a single‐line regimen until PD was detected. Overall survival (OS) was calculated from the first day of starting a single‐line regimen until death resulting from any cause or the last follow‐up visit. Survival time was estimated using the Kaplan‐Meier method and compared between the groups with the log‐rank test. The Cox proportional hazards model was applied for multivariate survival analyses. Two‐sided p‐values <.05 were considered significant.

Results

Clinical Characteristics of All Patients Analyzed

The demographic characteristics of the 246 evaluable patients who were categorized into MR and unmixed response (unMR) subgroups are summarized in Table 1. Patients with advanced NSCLC (IIIB‐IV) had a higher incidence of MR (30.0%, 48/160) than those with earlier disease (I‐IIIA) (5.8%, 5/86) (p < .001). Patients with EGFR mutant type (50.9% [27/53] versus 32.6% [63/19], p = .044), pathological confirmation of adenocarcinoma (86.8% [46/53] versus 75.1% [145/193], p = .071), and patients treated with EGFR‐TKIs (47.2% [25/53] versus 28.0% [54/193], p = .008) tended to produce an MR compared with patients with EGFR wild type (WT), those with non‐adenocarcinoma NSCLC, and those receiving chemotherapy. With stable controls (PR or stable disease) on other organs based on individual data, the rate of progression resulting in an MR was highest for bones, followed by lymph nodes, the CNS, intrapulmonary metastases, pleura, and adrenal glands (supplemental online Fig. 2).

Clinical Characteristics and Classification of the Patients with MRs

Further clinical characteristics of the 53 enrolled patients assessed as MR are described. In total, 50.9% (27/53) of the patients had EGFR mutant NSCLC. Of the 53 patients, 47.2% (25/53) received EGFR‐TKI and 52.8% (28/53) received chemotherapy. Patients eventually evaluated as MR were divided into two groups: those with stable or local progression disease coupled with fewer symptoms (symptom score ≤1) were defined as efficacious MRs (eMR, n = 26, 49.1%), and those with dramatic progression or showed systemic symptoms (symptom score >1) were defined as inefficacious MRs (ineMR, n = 27, 50.9%). Symptom score was quantified as previously described on the basis of six items: cough, hemoptysis, chest pain, fever, dyspnea, and metastatic lesion‐related symptom [12]. Scores 0, 1, and 2 were quantified in accordance with the asymptomatic status, stability of preexisting symptom, and deterioration of any preexisting or new symptom [13]. Among patients with efficacious MR, 65.4% maintained prior regimens with or without local intervention, whereas most patients (63.0%) in the inefficacious MR group switched to next‐line regimens, including chemotherapy or EGFR‐TKIs based on the gene profile of re‐biopsy (Fig. 1A; supplemental online Table 1).

Notably, those with efficacious MR showed significantly prolonged PFS (9.4 versus 3.8 months; p = .012) and OS (26.5 versus 9.5 months; p = .027) compared with those in the inefficacious MR group (Fig. 1B, 1C), manifesting the potential reason for classification as an MR and subsequent treatment in clinical practice. We also analyzed the role of genetic heterogeneity in MR. There were 14 (26.4%) patients with confirmed gene heterogeneity before and after evaluation as MR. Patients with confirmed gene heterogeneity showed relatively shorter median PFS (4.6 versus 6.5 months; p = .567) and OS (10.2 versus 24.2 months; p = .226) than the uncertain group; however, there was no significant difference, which may partly result from the small sample size in each group (Fig. 1D, 1F).

Intertumoral Heterogeneity in Patients with an MR

A treatment‐naïve male nonsmoker was diagnosed with primary lung adenocarcinoma (cT2bN3M1a, stage IV) and an EGFR exon 19 deletion. After taking gefitinib for 1 year, the primary pulmonary lesion responded partially (−43%), but the disease was evaluated as having locally progressed to new lesions on mediastinal lymph nodes (region 2; diameter, 1.8 × 1.5 cm) based on PET/CT. The patient was categorized into the efficacious MR group and continued TKI plus local radiotherapy as subsequent treatment (Fig. 2A).

This typical case of MR shows how to differentiate and select optimal regimens for these patients. We further analyzed the correlation between gene changes, pathological types, MR type, and corresponding therapies of the 53 enrolled patients. Considering safety, feasibility, necessity, and the patient’s approval, five patients that received repeated biopsies for complete pathological and genetic profiles were confirmed to be typical genetic heterogeneous cases (Table 2). Four patients with the EGFR mutant NSCLC received TKIs as treatment and developed potential resistant mechanisms (two patients with exon 20 T790M and two patients switched to EGFR‐WT). With TKIs as the targeted treatment, the primary and metastatic lesions from which the biopsy specimens were obtained were under control, whereas the other lesions without biopsies and with no gene or pathological information became enlarged or progressed remarkably. Interestingly, patient 1 received pemetrexed + platinum combined with cetuximab after becoming resistant to erlotinib as first‐line therapy, but the resistance mutation disappeared, indicating re‐sensitivity to TKIs after the “TKI holiday.”

Subsequent treatment after MR is another important clinical issue, particularly in those with gene heterogeneity. Three patients categorized as inefficacious MR changed to next‐line treatment based on gene changes, and two other patients assessed as efficacious MR continued their previous therapy plus local intervention. However, most of them showed unfavorable responses and short PFS from the subsequent treatment, indicating a poor prognosis in these patients, and further effective treatments were needed (Table 2).

The gene mutation landscapes of 246 patients were analyzed to further evaluate whether gene heterogeneity served as the driving force to the MR. Changes in the EGFR mutant gene profiles in the mixed and non‐MR groups are described in detail. Among patients with EGFR mutant NSCLC, those with an MR were more inclined to develop T790M‐resistant mutations (18% versus 4%, respectively), switch to EGFR‐WT (19% versus 15%, respectively), or activate secondary driver mutations (15% versus 9%, respectively) compared with those without an MR. In total, a higher incidence of EGFR mutation heterogeneity was observed in the MR group than in the unMR group (52% versus 28%, respectively) (Fig. 2B).

Survival and Multivariate Analyses

A total of 160 patients with stage IIIB‐IV NSCLC (48 MR and 112 unMR) were included in the survival analysis. OS was significantly shorter in the MR than in the unMR group (12.8 versus 16.1 months; p = .031; Fig. 3B), particularly in those treated with chemotherapy (8.5 versus 14.7 months; p = .018; Fig. 3F). MR patients also showed a greater decrease in PFS than the unMR group, but with marginal significance (6.2 versus 7.0 months; p = .06; Fig. 3A). The subgroup analysis showed that TKI treatment significantly increased the difference in PFS between the groups (9.2 versus 10.1 months; p = .04; Fig. 3C).

Multivariate analyses confirmed that MR (hazard ratio [HR], 1.474; 95% confidence interval [CI], 1.018–2.134; p = .040) and male patients (HR, 1.565; 95% CI, 1.005–2.439; p = .047) served as unfavorable prognostic factors for PFS. Remarkably, MR (HR, 1.849; 95% CI, 1.190–2.871; p = .006), EGFR‐WT (HR, 2.228; 95% CI, 1.133–4.381; p = .020), and smoking status (HR, 1.590; 95% CI, 1.020–2.475; p = .040) were three independent risk factors for OS. These results indicate that patients with an MR predicted unfavorable survival (Table 3).

Discussion

This is the first study to our knowledge to systemically assess patients by PET/CT and differentiate MR and therapies, with a relatively large sample size. The overall incidence of MR within unselected patients was 21.5% (53/246) and tended to occur in those with advanced lung adenocarcinoma treated with a TKI, indicating that MR was not a rare event in patients with NSCLC and resulted from selective pressures of targeted therapy and tumor evolution. This study determined that MR was an unfavorable prognostic factor for survival. Patients with MR had a shorter PFS and poorer OS, as reported previously [13], indicating the need to recognize these patients to improve their potentially poor survival.

Subsequent treatment after MR is a clinical concern. Whether the patient should switch to next‐line treatment, continue the original regimen, or be followed‐up frequently has perplexed oncologists. Based on National Comprehensive Cancer Network guidelines, for patients who have experienced disease progression either during or after first‐line therapy, single‐agent docetaxel, pemetrexed, or erlotinib are established as second‐line regimens. In this study, we divided these populations into efficacious MR and inefficacious MR based on treatment response and symptom score. The classification of efficacious and inefficacious MR is mostly aimed to identify the potential mechanisms of MR, how MR effects patients’ benefits, and what we should do to differentiate patients with inefficacious MR and thus enhance prophylactic interventions. Efficacious MR with a better outcome may probably benefit from a reasonable subsequent therapy, which supported our viewpoint that different classifications of MR deserve different treatment. Most patients with efficacious MR were administered prior regimens with or without local therapy, whereas most of the inefficacious MR patients switched to alternative regimens based on driver gene status. It was not appropriate to withdraw the current regimens of patients who had partial disease flare‐ups, despite assessments using the RECIST criteria [14]. Continuation of their regimen combined with local interventions should be optimal for those with an efficacious MR, based on experience with multidisciplinary treatment . There is a general rule that therapies are indeed guided by druggable genetic alterations derived from known resistant mechanisms. However, most diseases progress with unknown mechanisms, and published literatures considered that they may result from inter‐ and intratumoral heterogeneity of primary and metastatic lesions.

Our results indicate that lesions within a patient with MR that were responsive to the current treatment were chiefly those from which repeated biopsies were obtained. This might be because the subsequent therapies were set up based on biopsy sites with a more comprehensive genetic mutation landscape than those of other lesions, indicating a need for multiple serial biopsies [16]. Among all metastatic sites, the highest incidence of progressive lesions occurred in bone followed by lymph nodes and the CNS, which conflicts with a previous study . Although drug penetration might be another important reason, different populations with heterogeneous sub‐driver mutations act as initiating resources of heterogeneity, which suggests more emphasis on frequently progressive sites with local interventions. However, invasive biopsies are not possible at every anatomic site for safety and feasibility. Therefore, clinical predictors for identifying patients who would probably exhibit heterogeneous responses to current therapies are needed, and it is imperative to explore potential mechanisms to assist overcoming those for tailored therapies.

Tumor heterogeneity in MR to targeted regimens or even systemic chemotherapy to some extent reflected the heterogeneous biological characteristics of different lesions within the same patient, either temporally or spatially, as reported previously . Several studies have identified intertumoral heterogeneity between multiple pulmonary nodules and associated metastatic sites , despite controversies. In this study, five patients with typical MR who were evaluable for both imaging and gene profiles from repeated biopsies manifested intertumoral heterogeneity, which might have been induced by EGFR‐TKIs or chemotherapy . Selective pressure from targeted regimens plays an important role in reflecting and affecting tumor behaviors, thus fostering resistance to ongoing therapies. Acquired or primary EGFR T790M [23], amplification of c‐MET [24], aberrant activities of bypass pathways [4], abnormal downstream pathways such as KRAS mutations [3], histological transformation [25], and genetic changes to EGFR‐WT are all identified resistance mechanisms. On the other hand, spatial intertumoral heterogeneity suggests different features of each metastatic lesion, resulting from various epigenetic microenvironments. For example, CNS lesions tend not to respond to drugs due to inadequate penetration across the blood–brain barrier but also bear less molecular resistance and secondary mutations . Our previous study illustrated dynamic changes in plasma EGFR mutations along with administration and discontinuation of TKI‐targeted therapies [28], indicating that disease flare‐ups were possibly associated with re‐sensitivity to TKIs during a drug‐free period [5]. Circular targets involving rechallenges with TKIs have previously been addressed . Therefore, serial and multiple intertumoral biopsies would provide the whole gene landscape to detect heterogeneity.

There were some limitations to this study. Data were collected retrospectively according to the medical records system of a single institution. Incomplete imaging tests in the general population may have led to an underestimation of the incidence of MR. Therefore, eligible patients were limited to those evaluated by PET/CT to provide comprehensive responses of systemic lesions and minimize selective bias from routine tests such as CT and MRI. Second, only one or more evaluable periods of each patient were analyzed, but the effects of previous therapies on evaluable periods were not illustrated. In addition, only EGFR mutations in eligible patients were analyzed, instead of interpreting all other actionable mutations in genes such as Echinoderm Microtubule Associated Protein Like 4-Anaplastic Lymphoma Receptor Tyrosine Kinase (EML4-ALK), and Tyrosine-Protein Kinase Met (c-MET), or immune expression of PD‐1/PD‐L1. Actually, gene profiles of patients have been tested to provide full‐scale guidance for molecular therapies in routine clinical practice. The treatment strategies in this study did not interfere with the interpretation of the results.

Conclusion

This study found a relatively high incidence of MR in patients with advanced NSCLC, particularly those treated with targeted therapies. Clinical features, such as progressive organ type, might provide clues when multiple and serial biopsies are unavailable. An MR may be an unfavorable prognostic factor and originate from morphological and genetic intra‐/inter‐heterogeneity. Large prospective studies involving the entire genetic landscape of patients with NSCLC are imperative to explore mechanisms of MRs and guide tailored therapies to improve OS.

Acknowledgments

This study was supported by the Key Technologies Research and Development Program of Guangzhou (2011Y2‐00014), the Key Laboratory Program of Guangdong (2012A061400006), the Special Fund for Research in the Public Interest from the National Health and Family Planning Commission of the People’s Republic of China (Grant 201402031), and the Project of National Natural Science Foundation (Grant 81372285).

Author Contributions

Conception/Design: Yi‐Long Wu, Zhong‐Yi Dong, Wen‐Zhao Zhong

Provision of study material or patients: Hao‐Ran Zhai, Qing‐Yi Hou, Wen‐Zhao Zhong

Collection and/or assembly of data: Zhong‐Yi Dong, Qing‐Yi Hou, Yang‐Si Li, Zhi‐Yong Chen

Data analysis and interpretation: Zhong‐Yi Dong, Hao‐Ran Zhai, Jian Su, Hong‐Hong Yan, Wen‐Zhao Zhong

Manuscript writing: Zhong‐Yi Dong, Hao‐Ran Zhai, Si‐Yang Liu

Final approval of manuscript: Yi‐Long Wu, Zhong‐Yi Dong, Hao‐Ran Zhai, Qing‐Yi Hou, Jian Su, Si‐Yang Liu, Hong‐Hong Yan, Yang‐Si Li, Zhi‐Yong Chen, Wen‐Zhao Zhong

Disclosures

The authors indicated no financial relationships.

References

Author notes