Complete Metabolic Response on Interim ¹⁸F‐Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography to Predict Long‐Term Survival in Patients with Breast Cancer Undergoing Neoadjuvant Chemotherapy

The objective of this study was to determine the prognostic role of complete metabolic response on ¹⁸F‐fluorodeoxyglucose (¹⁸F‐FDG) positron emission tomography/computed tomography during or following neoadjuvant chemotherapy in breast cancer patients. The association between ¹⁸F‐FDG metabolic parameters (including uptake) and survival outcome, according to breast cancer subtype, was investigated

Abstract

Background.

This study aims to investigate the prognostic role of complete metabolic response (CMR) on interim ¹⁸F‐fluorodeoxyglucose positron emission tomography/computed tomography (PET/CT) in patients with breast cancer (BC) receiving neoadjuvant chemotherapy (NAC) according to tumor subtypes and PET timing.

Patients and Methods.

Eighty‐six consecutive patients with stage II/III BC who received PET/CT during or following NAC were included. Time‐dependent receiver operating characteristic analysis and Kaplan‐Meier analysis were used to determine correlation between metabolic parameters and survival outcomes.

Results.

The median follow‐up duration was 71 months. Maximum standardized uptake value (SUV_max) on an interim PET/CT independently correlated with survival by multivariate analysis (overall survival [OS]: hazard ratio: 1.139, 95% confidence interval: 1.058–1.226, p = .001). By taking PET timing into account, best association of SUV_max with survival was obtained on PET after two to three cycles of NAC (area under the curve [AUC]: 0.941 at 1 year after initiation of NAC) and PET after four to five (AUC: 0.871 at 4 years), while PET after six to eight cycles of NAC had less prognostic value. CMR was obtained in 62% of patients (23/37) with estrogen receptor‐positive (ER+)/human epidermal growth factor receptor 2‐negative (HER2−) BC, in 48% (12/25) triple‐negative BC (TNBC), and in 75% (18/24) HER2‐positive (HER2+) tumors. Patients with CMR on an early‐mid PET had 5‐year OS rates of 92% for ER+/HER2− tumors and 80% for TNBC, respectively. Among HER2+ subtype, 89% patients (16/18) with CMR had no relapse.

Conclusion.

CMR indicated a significantly better outcome in BC and may serve as a favorable imaging prognosticator. The Oncologist 2017;22:526–534

Implications for Practice.

This study shows a significantly better outcome for breast cancer (BC) patients who achieved complete metabolic response (CMR) on ¹⁸F‐fluorodeoxyglucose emission tomography/computed tomography (PET/CT) during neoadjuvant chemotherapy, especially for hormone receptor‐positive tumors and triple negative BC. Moreover, PET/CT performed during an early‐ or mid‐course neoadjuvant therapy is more predictive for long‐term survival outcome than a late PET/CT. These findings support that CMR may serve as a favorable imaging prognosticator for BC and has potential for application to daily clinical practice.

Introduction

Neoadjuvant chemotherapy (NAC) is commonly offered to patients with stage II/III breast cancer (BC) with the goal of delivering early systemic therapy to micrometastatic foci, allowing more patients to undergo breast‐conserving surgery, and assessing tumor response to systemic therapy agents in vivo [1]. Because pathological complete response (pCR) has been validated as a surrogate endpoint of survival in patients receiving NAC for BC, the most widely studied role of ¹⁸F‐fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F‐FDG PET/CT) for response assessment was to predict pCR. However, pCR is seldom attained in estrogen receptor‐positive (ER+) tumors, despite their favorable prognosis. Recent pooled analyses have shown that pCR is not associated with improved survival in less aggressive subtypes such as ER+ BC [2], [3].

¹⁸F‐FDG PET provides a reliable measure of tumor burden, as the amount of FDG uptake reflects the number of viable tumor cells. Residual tumor metabolic activity on PET after therapy is an indicator of active disease, whereas a midcourse negative PET is predictive of a successful treatment response [4], [5]. Complete metabolic response (CMR), defined as negative findings on a PET study during or following antitumor therapy, has been proven to predict survival in lymphoma [6], cervical cancer [7], and metastatic BC [8]. The prognostic value of CMR in patients with BC receiving NAC remains uncertain and requires further investigation.

In this study, we aimed to determine the prognostic role of CMR on ¹⁸F‐FDG PET/CT during or following NAC in BC patients. Our key objectives were to investigate the association between FDG uptake and survival outcome according to BC subtypes, to evaluate the prognostic performance of CMR with different standards, and to assess the impact of PET timing on the prognostic performance of FDG uptake.

Materials and Methods

Patients

The retrospective protocol was approved by the institutional review board, and informed consent was waived. A search of the PET/CT clinic database identified 218 consecutive patients with BC who underwent ¹⁸F‐FDG PET/CT for evaluation of treatment response to NAC between 2003 and 2011. To be eligible for our study, patients had to have histologically confirmed BC stage II or III, must have received NAC, and must have had at least one ¹⁸F‐FDG PET/CT scan during or following NAC. Patients all had received anthracycline‐based and/or taxane‐based regimens; trastuzumab was added, as indicated, when human epidermal growth factor receptor 2 (HER2) was overexpressed or amplified. Patients with ER+ tumors received at least 5 years of endocrine therapy after surgery, and patients with HER2+ tumors received 1 year of adjuvant anti‐HER2 therapy. In patients who had more than one interim PET/CT scans, the first PET/CT study was evaluated. Exclusion criteria are detailed in the flow chart (supplemental online Fig. 1).

For our analysis, pCR was defined as the absence of invasive BC in the breast and axillary nodes irrespective of ductal carcinoma in situ. Event‐free survival (EFS) and BC‐specific overall survival (OS) served as the main outcome measures. EFS was defined as the time from the first day of NAC to the day of disease progression (i.e., loco‐regional or systemic disease recurrence, distant metastasis, death from cancer, or unequivocal imaging evidence of progression). OS was defined as the time from the first day of NAC to the day of cancer‐related death. Because different subtypes of BC show different glucose metabolism and clinical outcomes, analyses were conducted first in the whole study population and then in each of the following BC subtypes: ER‐positive/HER2‐negative (ER+/HER2−), HER2‐positive (HER2+), and triple negative BC (TNBC).

PET/CT Acquisitions and FDG Uptake Measurements

¹⁸F‐FDG PET/CT was performed according to a standard clinical protocol. Briefly, the patients were required to fast for at least 6 hours, and blood glucose levels had to be lower than 200 mg/dL. Whole‐body, three‐dimensional PET/CT scans were acquired using a Discovery ST 8, STE 16, RX 16, or VCT 64 PET/CT scanner (GE Healthcare, Milwaukee, WI, http://www.gehealthcare.com) at approximately 60 minutes after intravenous administration of 259–555 MBq of ¹⁸F‐FDG. A CT scan (120 kV; 300 mA; pitch, 1.375; slice thickness, 3.75 mm; rotation time, 0.5 second) without contrast enhancement was obtained and used for attenuation correction and diagnostic purposes.

PET/CT images were fused and read on Advantage Workstation 2 (GE Healthcare). Three‐dimensional regions of interest were drawn around the primary tumors and around metastatic lymph nodes when present, then maximum standardized uptake values (SUV_max) were automatically generated. In 49 patients with a baseline PET/CT, the reduction of SUV_max was calculated as: ΔSUV_max = (SUV_max at baseline – SUV_max of the second PET)/SUV_max at baseline * 100%.

PET Timing

Four groups were divided according to their number of courses of NAC administered before PET acquisition: PET_2‐3c — patients underwent PET/CT studies after two to three cycles of NAC (n = 17); PET_4‐5c — PET/CT was performed after four to five cycles of NAC (n = 28); PET_6‐7c — PET/CT after six to seven cycles of NAC (n = 18); PET_8c — PET/CT scans after eight cycles of NAC (n = 23).

CMR Definitions

The following three definitions of CMR on PET/CT were compared: CMR_breast as — SUV_max lower than FDG uptake in contralateral breast tissue; CMR_liver as — SUV_max below liver FDG uptake; and CMR_suv as — SUV_max lower than a threshold of SUV_max, which is determined by receiver operating characteristic (ROC) curve analysis and Youden Index. CMR was described within the whole‐body PET/CT scan regardless of CT findings. Conversely, persistence of abnormal FDG uptake (SUV_max higher than the reference settings) was considered CMR not achieved (no‐CMR). Diffuse uptake in the spleen or bone marrow was considered a result of therapy and was not recorded as active disease.

Statistical Analysis

ROC curve analysis was used to define CMR_suv by evaluating the optimal cutoff value of SUV_max to predict disease relapse and cancer‐death. The time‐dependent ROC curve, which estimated the specificity and sensitivity as functions of time [9], was used to evaluate the performance of SUV_max at different timing in predicting disease progression or death. The association between long‐term survival and metabolic parameters was investigated using the univariate and multivariate Cox proportional hazards regression analyses. Kaplan‐Meier survival analysis and the log‐rank test were utilized to determine median follow‐up time and 5‐year EFS and OS rates. Negative predictive values (NPVs) and positive predictive values (PPVs) were calculated according to the standard definition [10]. The “survivalROC” package for performance assessment and comparison in the open source statistical software R 3.2.3 were used for the time‐dependent ROC analyses. All other statistical analyses were performed using SPSS 22.0 (IBM, Chicago, IL, http://www.ibm.com) software. A p value less than .05 based on two‐sided tests was considered statistically significant.

Results

Patient Characteristics

Eighty‐six consecutive patients receiving NAC for stage II or III BC were included (Table 1). Forty‐nine of them received a baseline PET/CT. Patients’ median age was 51 years old. The median follow‐up duration was 71 months (range, 8–118 months). Twenty‐seven patients had disease recurrence and 23 had died. All patients underwent adjuvant chest wall and/or axillary‐region radiation therapy, except one patient with TNBC who had tumor relapsed on the chest wall, confirmed by ultrasonography‐guided biopsy, 15 days after bilateral mastectomy and removal of left subclavian Port‐A‐Cath.

Table 1. Patient characteristics and univariate Cox regression analysis for EFS and OS in the whole study population.

^aSUV_max on PET during or following neoadjuvant chemotherapy.

^bSUV_max of patients with both a baseline PET and an interim PET.

Abbreviations: CI, confidence interval; CMR, complete metabolic response; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HR, hazard ratio; NA, not available; pCR, pathological complete response; PR, progesterone receptor; SUV_max, maximum standardized uptake value; TN, triple negative.

The Whole Population

In the whole population with mixed subtypes and PET timing, univariate analysis revealed that SUV_max, lymph node (N) stage, and TNBC were associated with both EFS and OS (Table 1). ER status correlated with OS. SUV_max had the most significant association with survival. After adjusting for other clinicopathological and metabolic parameters by multivariate Cox proportional hazards regression analysis, SUV_max, TNBC, and N stage remained significant for predicting EFS, and SUV_max and TNBC had independent predictive value for OS (Table 2).

Table 2. Multivariate Cox regression analysis for event‐free survival and overall survival in the whole study population.

Abbreviations: CI, confidence interval; HR, hazard ratio; N stage, lymph node stage; NA, not available; SUV_max, maximum standardized uptake value; TN, triple negative.

ROC analyses determined that the areas under the curve (AUCs) of SUV_max was 0.692 (p = .007) to predict disease‐related death. The optimal SUV_max cutoff value was 2.5, calculated according to Youden index. Thus, CMR_suv using a threshold of SUV_max of 2.5 was described as CMR_2.5. As shown in Table 3, 43 out of 53 patients with CMR_2.5 experienced no disease relapse. Ten patients (12%) achieved CMR_breast and none of them experienced disease progression or cancer‐related death. During the follow‐up duration, 30 out of 38 patients with CMR_liver remained event‐free. Each definition of CMR stratified high‐ or low‐risk patients effectively (Fig. 1). CMR_2.5 offered the strongest correlation with survival.

Table 3. Negative and positive predictive values of pCR and CMRs for overall survival according to tumor subtypes.

Abbreviations: CMR, complete metabolic response; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; NPV, negative predictive value; pCR, pathological complete response; PPV, positive predictive value; TN, triple negative.

Figure 1.

Kaplan‐Meier survival plots for event‐free survival and overall survival according to complete metabolic response (CMR) definitions: CMR_breast (A, B), CMR_liver (C, D), and CMR_2.5 (E, F). Log‐rank p values are provided.

In 49 patients with baseline PET studies, neither the reduction of SUV_max nor an absolute SUV_max during or following treatment predicted survival by univariate Cox regression analysis (Table 1), while CMR_2.5 had significant correlation with both EFS (hazard ratio [HR]: 4.374, 95% confidence interval [CI]: 1.331–14.374, p = .015) and OS (HR: 4.667, 95% CI: 1.247–17.472, p = .022).

PET Timing

Time‐dependent ROC curve analysis revealed that SUV_max on PET_2‐3c and PET_4‐5c outperformed SUV_max on PET_6‐7c and PET_8c to predict survival (Fig. 2). AUCs as a measure of OS prediction was the best with PET_2‐3c of 0.941 at 1 year after the initiation of cancer treatment, followed by PET_4‐5c of 0.871 at the fourth year. For EFS prediction, AUCs were the best with PET_2‐3c of 0.800 at the first year, followed by PET_4‐5c of 0.748 at the second year. Optimal cutoff values of SUV_max on PET_2‐3c and PET_4‐5c were 2.7 and 2.5, respectively, which was similar or the same as the cutoff value in the whole study population.

Figure 2.

Time‐dependent receiver operating characteristic curves as a function of time for maximum standardized uptake values (SUV_max) on positron emission tomography (PET)/computed tomography of different timing to predict event‐free survival (A) and overall survival (B): black dashed line, SUV_max at PET performed after two to three cycles of NAC; black solid line, SUV_max at PET performed after four to five cycles of NAC; grey dashed line, SUV_max at PET_6‐7c; grey solid line, SUV_max at PET_8c. Abbreviations: AUC, area under the curve.

CMR_2.5 on PET_2‐3c and PET_4‐5c resulted in a stronger stratification of patients at high‐ and low‐risk than a late PET (Fig. 3). Five‐year EFS and OS rates for patients with CMR_2.5 on PET_2‐3c were 86% versus 27% (p = .024) and 86% versus 56% (p = .140), respectively. Five‐year EFS and OS rates for patients with CMR_2.5 on PET_4‐5c were 85% versus 37.5% (p = .008) and 95% versus 37.5% (p <.001), respectively. No statistical significance in survival was observed between patients who achieved CMR_2.5 and those with no‐CMR_2.5 either on PET_6‐7c or PET_8c (Fig. 3E–3H).

Figure 3.

Kaplan‐Meier survival curves of complete metabolic response with a maximum standardized uptake value of 2.5 according to positron emission tomography (PET) timing: PET_2‐3c (A, B); PET_4‐5c (C, D); PET_6‐7c (E, F); PET_8c (G, H). Log‐rank p values are provided.

ER+/HER2− Tumors

Thirty‐seven women had an ER+/HER2− tumor, ten of them progressed, and eight had died at the time of analysis. Patients who achieved pCR (n = 1) or CMR_breast (n = 5) remained event‐free. In 36 patients with no‐pCR, 26 (72%) remained event‐free. Twenty‐three patients obtained CMR_2.5 and 17 had CMR_liver. NPVs and PPVs of pCR and CMR definitions for OS are provided in Table 3. For patients with CMR_2.5, 5‐year EFS rates were 81% versus 51% (p = .041) and 5‐year OS rates were 91% versus 67% (p = .027), respectively.

Due to the small number of patients with specific subtypes in each timing group, PET_2‐3c and PET_4‐5c were combined to form a group of early‐mid PET as PET_2‐5c. Similarly, PET_6‐7c and PET_8c were combined together to form a group of late PET as PET_6‐8c. As a result, 21 patients had a PET study after two to five cycles of NAC (PET_2‐5c), and 16 patients had a late PET (PET_6‐8c). ROC analyses revealed that SUV_max on PET_2‐5c had a higher AUC than SUV_max on PET_5‐8c for prediction of both disease progression (0.711 versus 0.490) and death (0.794 versus 0.490). Correlation between a residual SUV_max at different PET timing and survival in this subtype is shown in Table 4. Higher SUV_max on an early‐mid PET study indicated inferior survival. CMR_2.5 on PET_2‐5c demonstrated a clearer distinction between high‐ and low‐risk patients with 5‐year EFS of 85% versus 38% (p = .023) and 5‐year OS of 92% versus 64% (p = .023), respectively.

Table 4. Univariate Cox regression analysis for SUV_max on PET of different timing to predict event‐free survival and overall survival according to tumor subtypes.

^aDue to limited number of events, univariate analysis according to PET timing was not performed.

Abbreviations: CI, confidence interval; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HR, hazard ratio; SUV_max, maximum standardized uptake value; TN, triple negative.

TNBC.

In the group of 25 TNBC, 13 patients had relapse events and 12 died. Three out of four patients with pCR remained event‐free. NPVs and PPVs of each CMR definition are provided in Table 3. Five‐year OS rates for patients with CMR_2.5 were 73% versus 31% (p = .037).

Taking PET timing into account, ROC analyses showed that SUV_max on PET_2‐5c had a higher AUC than SUV_max on PET_5‐8c for prediction of death (0.816 versus 0.667) and a slightly better AUC to predict disease progression (0.687 versus 0.667). As shown in Table 4, SUV_max on an early‐mid PET had a stronger association with OS (HR: 1.155, 95% CI: 1.007–1.324, p = .040), as compared with a late PET (HR: 1.030, 95% CI: 0.882–1.204, p = .707). Although statistical significance was not reached, due to the small number of patients, 5‐year OS of patients with CMR_2.5 on PET_2‐5c was 80% versus 33% (p = .133). In group of PET_6‐8c, 5‐year OS rates of CMR_2.5 was 60% versus 40% (p = .568).

HER2+ Subtype

Twenty‐four patients had HER2+ tumors. Four patients relapsed and three died. During follow‐up, eight out of ten patients with pCR had no disease‐related death. Sixteen out of 18 patients with CMR_2.5 had no relapse (Table 3). A residual SUV_max had no significant correlation with either EFS or OS. Although nonsignificant, a trend towards longer survival in HER2+ patients who achieved CMR_2.5 was observed, with 5‐year EFS rates of 88% versus 67% (p = .197). Given the small number of relapse and death events, analyses according to PET timing was not performed.

Discussion

With a median follow‐up duration of 71 months, FDG uptake on a single PET/CT independently correlated with survival outcomes by multivariate analysis in BC patients receiving NAC. CMR during or following NAC predicted a favorable outcome. This study also observed that an early or midcourse PET/CT had a higher predictive value for survival compared with a late PET.

CMR during or following chemotherapy can powerfully stratify for survival in patients with metastatic BC [8], [11]. In BC patients with mixed tumor subtypes receiving NAC, Emmering et al. [12] presented that failure to achieve a negative preoperative PET indicated 2.5 times higher risk of disease recurrence. In this study, we first compared the prognostic value of three different definitions of CMR in the whole cohort, using the contralateral breast tissue, liver background, or a cutoff value of SUV as reference settings. We found each CMR definition had a role in prognosis. CMR_breast predicted an excellent prognosis (NPV of 100%), and CMR_2.5 stratified for survival most powerfully (Fig. 1).

Because BC with different subtypes has different glucose metabolism, treatment response, and clinical outcomes [13], [14], we aimed to investigate the prognostic role of residual FDG uptake and CMR during or following NAC according to tumor subtypes and PET timing. In ER+/HER2− BC, previous studies have proven the prognostic value of change in SUV_max after 1 or 2 courses of NAC [15], [16], [17]. However, the cutoff values of SUV reduction varied greatly across these studies, from 12% to 50%, making interpretation of scan results challenging in daily clinical practice. Comparing with change in SUV_max, Groheux et al. [16] found that an absolute SUV_max after two cycles of NAC had a stronger association with survival and adversely affected prognosis for ER+/HER2− BC. Our data are in agreement and showed SUV_max and CMR during or following NAC significantly correlated with long‐term survival in this subtype. As opposed to low rate of pCR (3%), CMR_2.5 was achieved at a frequency of 62% in ER+/HER2− tumors. After taking PET timing into account, both SUV_max and CMR_2.5 on an early‐mid PET showed higher predictive value than a late PET in this subtype (Table 4). Thus, SUV_max and CMR_2.5 on an early‐mid PET might serve as imaging prognostic factors in ER+/HER2− BC.

In aggressive tumor subtypes (TNBC or HER2+), our findings on the prognostic role of PET parameter confirm and extend on previous reports. Despite contradicting results, the ability of SUV reduction to predict pCR [18], [19], [20], [21] and survival [22], [23], [24] has been investigated in TNBC and HER2+ tumors. However, limited information is available on the relationship between survival and residual FDG uptake during or following NAC in these subtypes. In this study, an absolute residual SUV_max showed only marginal or no relation with survival for TNBC or HER2+ tumors. After grouping by PET timing, however, higher SUV_max on PET_2‐5c suggested inferior OS in TNBC. Although no conclusion was drawn regarding CMR, a table in a study by Groheux et al. [24] containing 20 TNBC patients recorded that five out of six (83%) patients with an absolute SUV_max below 2.5 after two cycles of NAC had no relapse. Our results are in agreement; approximately 80% of TNBC patients who achieved an early CMR_2.5 during NAC did not die of cancer. We also found evidence on the prognostic role of CMR_2.5 in the HER2+ subtype, showing near 90% of patients with CMR experienced no relapse.

Our study adds to the accumulating evidence that the correlation between metabolic parameters of PET and survival varied by PET timing. Several studies compared PET performance at different time points and found that an early assessment after one or two cycles of treatment is more predictive [25], whereas others proved that PET at midcourse [19], [26] or prior to surgery might be optimal [27]. This study differed from the previous studies, which focused on the role of change in FDG uptake between sequential PET scans in patients with mixed subtypes. We assessed the impact of PET timing on the prognostic performance of SUV on a single interim or preoperative PET according to BC subtypes. Time‐dependent ROC analysis allowed comparison of observations at different time points [9] and revealed that AUC as a measure of EFS or OS prediction was the highest with PET_2‐3c by the first 1 or 2 years, while it was much lower at later follow‐up times. PET_4‐5c, however, appeared to have a higher AUC for predicting both survival outcomes at later time points (Fig. 2). Thus, FDG uptake on PET after two to three cycles of NAC might be more accurate at identifying patients who relapse or die within the first 1 or 2 years, while PET after four to five cycles of NAC appears to provide improved discrimination for survival at later follow‐up times. PET performed after six to eight cycles of NAC was less prognostic. Furthermore, CMR on an early or midcourse PET more powerfully stratified for survival both in the whole population and in groups of BC subtypes. The results are in concordance with the theory of “the kinetics of tumor cell killing and relation to PET” by Kasamon et al. [5], which has been widely implemented on treatment response assessment in lymphoma. They indicated that a midcourse negative PET study might represent an absence of cancer cells or minimal residual disease that would lead to cure after chemotherapy, while a negative scan later or at the end of therapy may have less value in distinguishing between an absence of tumor and an insufficient tumor cell killing required to achieve a cure. In our study, a residual FDG uptake at a late PET failed to predict survival both in the whole cohort and among different tumor subtypes (Fig. 3).

Our study has limitations, mostly related to its retrospective nature. First, the sample size was small and the patient population was heterogeneous. Despite the small number, our cohort had representative distribution with subtype distribution and pCR frequency similar to other reports [2], [19], [22]. Second, PET timing was variable. Due to heterogeneous PET timing and limited number of patients who had received a baseline PET in our study, the correlation between survival and the reduction of SUV_max or residual FDG uptake during or following NAC could not be validated. Despite this, after taking account of PET timing, our findings confirmed the theory of “the kinetics of tumor cell killing and relation to PET,” showing that FDG uptake and CMR on an early or midcourse PET might reflect tumor sensitivity to certain regimens of NAC and had higher prognostic value than a late PET. An additional limitation was that the ROC analysis and Youden index were used to define the cutoff value of SUV_max 2.5 as CMR_suv, and this method could overestimate the predictive performance of CMR_2.5. Hence, CMR_liver and/or CMR_breast should be interpreted as well, especially in aggressive subtypes such as TNBC. Prospective studies with larger sample sizes are needed to confirm the findings.

Conclusion

In summary, our findings suggest that FDG uptake on a single PET/CT after initiation of NAC was an independent prognostic factor for long‐term survival in BC patients. An early or midcourse CMR indicated an excellent prognosis, with 5‐year OS rates of 92% for ER+/HER2− tumor and 80% for TNBC, respectively.

See http://www.TheOncologist.com for supplemental material available online.

Author Contributions

Concept/Design: Suyun Chen, Nuhad K. Ibrahim, Yuanqing Yan, Stephen T. Wong, Hui Wang, Franklin C. Wong

Provision of study material or patients: Suyun Chen, Nuhad K. Ibrahim, Hui Wang, Franklin C. Wong

Collection and/or assembly of data: Suyun Chen, Nuhad K. Ibrahim, Yuanqing Yan, Stephen T. Wong, Franklin C. Wong

Data analysis and interpretation: Suyun Chen, Nuhad K. Ibrahim, Yuanqing Yan, Stephen T. Wong, Hui Wang, Franklin C. Wong

Manuscript writing: Suyun Chen, Nuhad K. Ibrahim, Yuanqing Yan, Stephen T. Wong, Hui Wang, Franklin C. Wong

Final approval of manuscript: Suyun Chen, Nuhad K. Ibrahim, Yuanqing Yan, Stephen T. Wong, Hui Wang, Franklin C. Wong

Disclosures

The authors indicated no financial relationships.