A, Venn diagram of patterns of failure. B, Distribution of treatment failure over time.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hosni A, Huang SH, Xu W, et al. Distant Metastases Following Postoperative Intensity-Modulated Radiotherapy for Oral Cavity Squamous Cell Carcinoma. JAMA Otolaryngol Head Neck Surg. 2017;143(4):368–375. doi:10.1001/jamaoto.2016.3668
What are the characteristics and risk factors of distant metastases (DM) in oral squamous cell carcinoma following postoperative intensity-modulated radiotherapy?
In this retrospective study, with prospectively acquired data, the clinicopathological characteristics of distant-only failure and DM with locoregional failure were similar. Both pN2-3 and G2-3 were independent predictors of DM.
Our results can help in the identification of a subset of high-risk patients characterized by poor clinical outcomes owing to high metastatic potential with a short latency of metastatic disease. These factors could also be employed for trials of treatment intensification or screening strategies for DM in the early years after surgery.
Advances in surgical techniques, the advent of intensity-modulated radiotherapy (IMRT), and the use of concurrent chemotherapy in oral squamous cell carcinoma (OSCC) have led to improvement of locoregional control (LRC), but not distant control (DC). Moreover, the development of distant metastases (DM) in OSCC has a dismal prognosis.
To determine the characteristics and risk factors of DM following postoperative IMRT in OSCC, and to identify the clinicopathological features that could be associated with distant-only failure (DOF).
Design, Setting, and Participants
Retrospective study of 300 OSCC patients (192 [64%] men and 108 [36%] women) treated with surgery and postoperative IMRT between 2005-2012 in a tertiary cancer center.
All patients underwent initial primary curative-intent resection with postoperative IMRT with or without concurrent chemotherapy based on predefined risk features.
Main Outcomes and Measures
Locoregional control, DC, overall survival (OS), and Radiation Therapy Oncology Group grade of 3 or higher late toxic effects. Multivariable analysis identified predictors for DM.
Overall 300 patients were identified (histological grade 2-3 [G2-3], 285 [95%]; pT3-4, 121 [41%]; pN2-3, 141 [47%]). Positive resection margin was present in 64 of 300 (21%) patients and extracapsular extension in 89 of 281 (32%) neck dissections. Median IMRT dose was 66 Gy and concurrent chemotherapy was used in 73 patients (24%). Median follow-up was 41 months. The 5-year local, regional, and distant control and OS were 85%, 82%, 86%, and 69%, respectively. On multivariable analysis, pN2-3 (hazards ratio, 5.7; 95% CI, 2.2-14.7) and G2-3 (HR, 4.9; 95% CI, 2.8-8.9) were predictive of DM. Thirty-nine patients developed DM, of which 20 (51%) were DOF and 12 (31%) were oligometastatic (≤5 lesions). The clinicopathological characteristics in DOF were similar to patients with DM subsequent to locoregional failure. In patients with G2-3, pN2-3, and extracapsullar extension (all together), the 5-year cumulative incidence of DOF was 22%.
Conclusions and Relevance
Surgery and postoperative IMRT with or without concurrent chemotherapy achieved encouraging outcomes. The clinicopathological characteristics of DOF and DM with locoregional failure were similar. Patients with G2-3, pN2-3, and extracapsullar extension (all together) have higher risk of DOF. Both pN2-3 and G2-3 were independent predictors of DM. Patients with these risk factors may be candidates for prospective clinical trials of intensified therapy or surveillance strategies.
The outcomes of oral cavity squamous cell carcinoma (OSCC) are typically worse than those at other common sites of head and neck squamous cell carcinoma (HNSCC). This is despite therapeutic advances, including modern surgical and radiation techniques and the use of postoperative concurrent chemotherapy in high-risk OSCC; however, distant metastases (DM) remain a challenge.1-3
In clinical practice, early detection of locoregional failure (LRF) in OSCC creates the chance for successful salvage treatment.4 On the other hand, neither routine clinical examination nor a tumor marker can provide early detection of DM in OSCC. Furthermore, the development of DM in OSCC has a poor prognosis; the addition of cetuximab to standard platinum-based chemotherapy (which represents the most significant advance in systemic treatment of DM in HNSCC) was associated with a median overall survival (OS) benefit of only 2.7 months (from 7.4 to 10.1 months).5 Therefore, understanding the characteristics of DM in OSCC and identifying risk factors for developing DM could improve early detection of DM, the likelihood of successful local ablative therapy of oligometastatic disease and, potentially, survival.6
Most of the studies on DM in HNSCC include analyses of mixed population of oral and nonoral SCC, without detailed reporting of the characteristics of distant-only failure (DOF) and DM subsequent to LRF. We have previously reported the patterns of LRF in patients with OSCC treated at our institution with postoperative intensity-modulated radiotherapy (IMRT).7 The rationale for conducting this study was to determine the characteristics and risk factors of DM following postoperative IMRT in OSCC, and to identify the clinicopathological characteristics that could be associated with DOF.
After institutional research ethics board approval from the Princess Margaret Cancer Center, a retrospective review was conducted of patients with newly diagnosed nonmetastatic (M0) OSCC, treated with curative intent at our institution between 2005 and 2012 with surgery and postoperative IMRT with or without concurrent (cisplatin) chemotherapy. Patients younger than 18 years and those with persistent or recurrent disease prior to initiation of postoperative IMRT were excluded from this analysis.
Clinical information including outcomes was retrieved from the Head and Neck Anthology of Outcomes where clinical and outcome data were prospectively collected at the point-of-care using the Formatted Anthology Synoptic Tick Sheet (FAST) process.8 Patients who had DM as the first site of treatment failure with no evidence of LRF until their last contact were reported as DOF, whereas the presence of DM after or simultaneously with LRF was recorded as subsequent DM. Distant metastases with 5 or fewer metastatic lesions in 1 organ were considered oligometatstic.
All patients underwent initial primary curative-intent resection. Cervical lymph node dissection (LND) was performed therapeutically in node-positive disease (cN+) or electively for clinically node-negative (cN0) cases if the risk for occult cervical metastases was estimated to be greater than 20% based on site, size, and thickness of the primary tumor. Bilateral LND was performed for cN2c disease or if there was a concern for bilateral lymphatic spread based on primary tumor location.
Postoperative radiotherapy (PORT) was determined clinically considering the following risk factors: pT3-4, pN2-3, level IV to V cervical lymph node involvement, nodal extracapsular extension (ECE), positive or close (<5 mm) microscopic margin(s), high grade, lymphovascular invasion (LVI), and perineural invasion (PNI). Concurrent chemotherapy (cisplatin) was administered in appropriate patients with positive margin(s) and/or ECE.
Radiological, pathological, and operative information was used by the radiation oncologist to delineate target volume. For patients with positive margin(s) and/or ECE, the high-risk clinical target volume (CTV1) included the site of preoperative gross disease (primary or nodal) with 5- to 10-mm margin. The operative bed and dissected nodal regions were delineated as an intermediate risk clinical target volume (CTV2). The low-risk clinical target volume (CTV3) included undissected, at-risk nodal regions.
CTV1, CTV2, and CTV3 plus an additional circumferential 5-mm margin (to accommodate the daily uncertainty) were delineated as PTV1, PTV2, and PTV3 respectively. Patients were treated with IMRT and routinely received a radiation dose of 66 Gy/33 fractions to PTV1, 60 Gy/30-33 fractions to PTV2, and 54-56 Gy/30-33 fractions to PTV3.
Patients were initially staged using computed tomographic (CT) scans of the head and neck with or without magnetic resonance imaging (MRI). A CT scan of the chest was performed for all except 10 patients with stages I to II disease. In patients with no smoking history, chest x-ray was deemed sufficient. Positron emission tomography (PET)-CT scan was not routinely used during the study period. Patients were typically seen in a multidisciplinary clinic with a comprehensive head and neck examination 2 to 6 weeks following PORT, every 3 months for the first 2 years, every 4 to 6 months in the third to fifth year, then annually if desired. Posttreatment imaging evaluation was done 10 to 12 weeks after the end of PORT, then as clinically indicated. Diagnosis of DM was based on follow-up imaging which was performed following the diagnosis of LRF or on the basis of new symptoms, while pathologic confirmation of DM was obtained in patients with limited disease in the lung only when it was difficult to differentiate radiologically between lung metastasis and primary lung cancer. Severe late RT-related toxic effects (LT) was defined as late Radiation Therapy Oncology Group (RTOG) grade of 3 or higher toxic effects starting more than 3 months after end of PORT.
Descriptive statistics were reported as median and range (or mean [SD]) for continuous variables, and as frequencies and proportions for categorical variables. The actuarial rates of local control (LC), regional control (RC), and distant control (DC) were estimated by the competing risk method. Overall survival was calculated with the Kaplan-Meier method. The estimated rate of DOF was assessed by the cumulative incidence function. Multivariable analysis using the Cox proportional hazards model was applied to identify predictors for DM and OS. Two-sided tests were applied. Effect size (Δ), and where appropriate, 95% CIs around the effect size were used. All the statistical analyses were conducted using SAS statistical software (version 9.4, SAS Institute Inc) and R (versions 3.1.3-3.3.1, R Foundation).
A total of 300 eligible patients were identified; 192 (64%) men and 108 (36%) women. The most common primary site (135 [45%] patients) was oral tongue. The median age at diagnosis was 61 years (range, 21-87 years) and the median follow-up for surviving patients was 41 months (range, 4-115 months). No specific clinicopathologic features were associated with DOF vs subsequent DM. However, compared with patients who achieved tumor control (ie, no LRF and no DM), patients with DOF more often had advanced nodal category (pN2-3, 85% vs 37% [Δ 48%; 95% CI, 31%-65%]) and ECE (55% vs 25% [Δ, 31%; 95% CI, 8%-54%]) (Table 1).
All patients underwent curative resection. Negative surgical margins of 5 mm or more were achieved in 62 patients (21%), close margins of 1 to less than 5 mm in 99 (33%) patients, very close margins of less than 1 mm in 75 (25%) patients, and positive microscopic margin(s) in 64 (21%) patients.
LND was performed in 281 (94%) patients: unilateral in 177 (59%) and bilateral in 104 (35%), with pathologic nodal disease (pN+) present in 195 (69%) LND and ECE in 89 (46%) pN+. Of the 19 patients who did not undergo LND, 6 had T1 disease (2 tongue, 1 buccal, 1 retromolar, 1 hard palate, and 1 lower lip), and 11 had T2 disease (4 tongue, 2 buccal, 2 alveolus, 1 retromolar, 1 floor of mouth, and 1 lower lip) and 2 had greater than T2 disease in maxillary alveolus, and maxillectomy without LND was performed at the surgeon’s discretion.
The median IMRT dose was 66 Gy in 33 fractions (range, 48-70 Gy); 156 (52%) patients received more than 60 Gy for positive and/or very close less than 1-mm margin(s) and/or ECE, while 8 (3%) patients received less than 60 Gy either owing to poor compliance with RT (6 patients) or early low risk (pT1N0) disease in elderly patients (2 patients, 79 and 91 years). One patient was treated with an accelerated hypofractionation schedule of 50 Gy in 20 fractions, and another with a 6-week hyperfractionation schedule (59.4 Gy in 54 fractions, 1.1 Gy per fraction, twice daily) owing to a previous medical history of head and neck irradiation for oropharyngeal cancer of which he had been cured 14 years prior to diagnosis of his OSCC. Bilateral neck irradiation was used in 185 (62%) patients, while 90 (30%) patients were treated with unilateral neck irradiation, and 25 (8%) received radiotherapy to primary site only.
Among 131 patients with positive resection margin(s) and/or ECE, concurrent cisplatin was used in 73 (56%) patients, while the remaining 58 patients either refused to receive concurrent cisplatin (4 of 58, 7%) or were unsuitable for chemotherapy (elderly patients 70 years or older [14 of 58, 24%], comorbidity [15 of 58, 26%], poor postoperative performance status [4 of 58, 7%], or combination of 1 or more of these causes [21 of 58, 36%]). The concurrent chemotherapeutic regimen was cisplatin 100 mg/m2 every 3 weeks (55 patients), 75 mg/m2 every 3 weeks (9 patients), and 40 mg/m2 weekly (9 patients). Among the 64 patients who were scheduled to received cisplatin (75 or 100 mg/m2) every 3 weeks; only 10 (15.5%) were able to receive 3 cycles, 49 (76.5%) received 2 cycles, while 5 (8%) patients tolerated only 1 cycle. Cumulative cisplatin dose of 200 mg/m2 or more was given to 60 (82%) of the 73 patients who received concurrent chemotherapy.
The actuarial 2-year (5-year) LC and RC were 89% (85%) and 84% (82%) respectively (Figure 1). Local failure (LF) was recorded in 39 patients at a median time of 12 months (range, 4-63 months), while regional failure (RF) occurred in 49 patients at a median time of 10 months (range, 2-39 months). Most LFs (30 of 39, 77%) and RFs (45 of 49, 92%) occurred in the first 2 years from the date of surgery. Of 22 patients with both LF and RF, 20 had synchronous LRF, 1 developed LF first and 1 had initial RF (Figure 2).
Regional failure occurred in 37 (20%) of 185 patients treated with bilateral neck irradiation, 10 (11%) of 90 patients received unilateral neck irradiation, and 2 (8%) of 25 patients treated with primary site–only radiotherapy. Among 10 patients with RF after unilateral neck irradiation, 3 failed in the contralateral neck (in level IIb, Vb, and VIb).
The estimated 2-year and 5-year DC were 87% and 86% respectively (Figure 1). A total of 39 patients developed DM at a median (range) time of 12 (3-49) months. The most common site for DM was the lung (31 of 39 DM, 79%): 21 (68%) of 31 as lung-only DM, and 10 (32%) of 31 with DM in lung and other organs. The distribution and pattern of DM is summarized in Table 2.
In 20 patients, DOF were reported at a median (range) of 15 (3-49) months from the date of surgery, with 16 (80%) of them developing DOF within 2 years following the surgery, while 19 patients had DM subsequent to LRF at a median (range) of 9 (3-42) months from the date of surgery and 1 (0-26) months from the time of LRF. Pathologic confirmation of DOF in the lung was obtained in 4 patients. The distribution of DOF and subsequent DM over time is illustrated in Figure 2B.
Univariable analysis showed that concurrent chemotherapy (HR, 0.5; 95% CI, 0.3-0.9) and G1 (HR, 0; 95% CI, 0-0) were significantly associated with less development of DM, while pN2-3 (hazards ratio [HR], 8.7; 95% CI, 3.5-22.1), and ECE (HR, 3.7; 95% CI, 1.9-6.9) were associated with DM. The association between DM and pT3-4 category (HR, 1.06; 95% CI, 0.56-2.01), tumor thickness (HR, 1.02; 95% CI, 0.77-1.35), LVI (HR, 1.56; 95% CI, 0.7-3.49), PNI (HR, 1.35; 95% CI, 0.7-2.6), and positive surgical margin(s) (HR, 1.18; 95% CI, 0.56-2.48) were not significant. On multivariable analysis, only pN2-3 (HR, 5.7; 95% CI, 2.2-14.7) and G2-3 (HR, 4.9; 95% CI, 2.8-8.9) predicted for DM, while ECE was marginally associated with development of DM (HR, 2.0; 95% CI, 0.96-4.1) (Table 3).
On univariable analysis, G1 compared with G2-3 was associated with reduction in the development of DOF (HR, 0; 95% CI, 0-0), while pN2-3 (HR, 6.9; 95% CI, 2.1-23.4) and ECE (HR, 2.9; 95% CI, 1.21-6.95) were associated with DOF. Subsequently we estimated the cumulative incidence of DOF according to the number of these risk factors; and the rate of DOF at 5 years was 0% for no risk factors, 3% for 1 factor, 10% for 2 factors, and 22% for 3 risk factors.
Metastatectomy was performed in only 1 patient with oligometastasis (single-lesion lung metastasis). Chemotherapy was attempted in 4 patients with lung metastases (1 with DOF [received cisplatin-fluorouracil-cetuximab] and 3 with subsequent-DM [2 received docetaxel and 1 had cisplatin-fluorouracil]). One patient with lung metastasis, and 2 with bone metastases received palliative radiotherapy (20 Gy, 5 fractions per week). Other patients with DM (32 [82%] of 39) were treated with best supportive care (BSC) measures, unlike 47 patients with LRF without DM, 16 (34%) of whom were treated with salvage surgery, 4 (9%) with reirradiation, 3 (6%) with chemotherapy, and 24 (51%) patients were treated with BSC.
The 2- and 5-year estimates of OS were 75% and 69% respectively (Figure 1). Of 90 deaths, 58 (64%) were OSCC related, 5 (6%) related to a second cancer (non-HNSCC), 6 (7%) died from other health related (noncancer) causes, while there were 21 patients (23%) with unknown cause of death.
The median survival after the detection of recurrence for patients with LRF and DM were 4 months (0-73) and 3 months (0-44) respectively. Thirty (64%) of the 47 patients with LRF without DM, 14 (70%) of the 20 patients with DOF, and 17 (89%) of the 19 patients with subsequent DM, died within 1 year from the time of recurrence. The median survival following detection of DM for patients with oligometastases and diffuse metastases were 3 months (range, 0-7) and 1.5 months (range, 1-2), respectively. On univariable analysis, pT3-4 (HR, 1.63; 95% CI, 1.08-2.47), pN2-3 (HR, 2.02; 95% CI, 1.32-3.08), ECE (HR, 2.23; 95% CI, 1.46-3.4), and G2-3 (HR, 3.53; 95% CI, 0.84-14.75) correlated with OS; however, none of these variables predicted for OS on multivariable analysis (Table 3).
No grade 4 or 5 LT was reported. Grade 3 LT occurred in 31 (10%) patients including osteoradionecrosis in 16 (5%), dysphagia in 6 (2%), neck fibrosis in 6 (2%), and trismus in 3 (1%). Of 73 patients who had concurrent chemotherapy, 55 (75%) had prophylactic percutaneous endoscopic gastrostomy (PEG) tubes placed per institutional protocol; among these patients, the PEG tube was removed within 3 months of completing PORT in 34 (62%) patients, between 3 to 6 months in 10 (18%) patients, between 6 to 8 months in 7 (13%) patients, and 4 (7%) patients remained PEG tube dependent for more than 1 year at last follow-up. Among patients with PEG removal, the median time to removal was 2 months (range, 1-8 months) following PORT.
In our current analysis of OSCC, we show encouraging LC, RC, DC, and OS with postoperative IMRT with or without concurrent chemotherapy according to risk factors with limited late toxic effects. These results are in concordance with the findings of large prospective trials1-3 using similar approaches with conventional radiation techniques, and several other reports9-11 on OSCC cohorts treated with postoperative IMRT.
A trend toward reversed pattern of failure in OSCC has been observed following improvement of LRC, but not DC, with new surgical techniques and advent of IMRT.12 Furthermore, the addition of chemotherapy (concurrent cisplatin) to PORT has shown a 48% risk reduction of LRF in patients with high-risk features (positive margin and/or ECE) without a corresponding benefit in DC.1-3,13 Similarly, our multivariable analysis showed that concurrent chemotherapy (cisplatin) was not associated with better DC.
The compliance to concurrent chemotherapy remains a challenge. In our study, 46% of patients with some high-risk features did not receive cisplatin, compared with 2% in the RTOG 9501 study and 12% in the EORT 22931 study. Moreover, only 15.5% of patients in our study were able to receive all the 3 planned cycles of cisplatin compared with 61% and 49% in the RTOG 9501 and EORTC 22931 studies, respectively.1-3,13 Ang et al14 reviewed the compliance levels in several published phase 3 randomized head and neck trials and substantial fraction of patients from these trials did not receive the planned 3 cisplatin cycles, and a cumulative dose of 200 mg/m2 was suggested to be sufficient. In an analysis15 of 6 phase 3 radiotherapy trials, higher cumulative cisplatin dose was significantly associated with OS benefit; however, the impact of cumulative cisplatin dose on outcomes in adjuvant setting remains questionable. In our study, 82% of patients who had concurrent chemotherapy received a cumulative cisplatin dose of 200 mg/m2 or more; however, the limited number of patients (73) who received chemotherapy in our study did not allow further exploration of this factor.
The overall incidence of DM in OSCC (approximately 10%) is relatively lower compared with other HNSCC sites,16 and screening for DM in all OSCC patients is not routinely implemented in clinical practice. There is potential for benefit of secondary screening for DM in identified high-risk patients. Risk factors for DM found in published OSCC series included: histological grade; pN-category; number, size, and location of positive lymph node(s); ECE; pT-category; and age.16-22 In our results, the 5-year DC was 86% with most DM (34 [87%] of 39 patients) occurring within 2 years following the surgery (Figure 2B), highlighting the importance of follow-up and surveillance in this time period. In patients with G2-3, pN2-3, and ECE (all together), the incidence of DOF was 22% at 5 years, respectively. In addition, G2-3 and pN2-3 were independent predictors of DM. Although these data are suggestive only, prospective evaluation of an imaging-based DM screening regimen based on the timing and risk factors for DM identified herein could be of benefit for OSCC patients.
Management of DM in OSCC patients is complex and treatment choices have usually included BSC or palliative systemic therapy. In our study, 82% of patients with DM were treated with BSC only. Other retrospective studies reported that approximately one-third to one-half of the patients with DM were treated only with BSC19,23; however, the comparison of the treatment methods used in this study for DM with those of other institutions is fairly difficult because the management strategy is not dependent on institutional guidelines rather than other factors including, but not limited to, patient’s performance status, comorbidity, location and extent of metastatic disease, prior treatment, symptoms, patient preference, logistics, and availability of phase 1 and 2 trials where the treatment takes place.
The introduction of ablative therapy for oligometastatic DM in personalized cancer management, treatment options could translate into better outcomes. Meta-analysis of salvage surgery for pulmonary metastases from treated HNSCC (including oral and nonoral SCC) reported a 5-year OS of 29% following metastatectomy.24 Stereotactic body radiotherapy (SBRT) yielded a similar survival rate, which is comparable to less than 5% with BSC and less than 15% with systemic treatment.6 Interestingly, we found that approximately one-third of DM was oligometastatic; however, there was no significant difference between OS following DM between patients with oligometastatic vs diffuse DM, indicating that the definition of oligometastatic disease (≤5 lesions in single organ), despite being widely accepted, may not be applicable in OSCC. To date, it is unclear how to select oligometastatic patients who will benefit from local ablative treatment, and better knowledge of the distinct patterns of OSCC metastatic progression will be necessary to customize metastatic disease treatments. In this cohort, only 1 patient with oligometastasis who had DM after 2010 (when we have an established oligometastses protocol for DM from HNSCC) was treated with metastatectomy. Furthermore, 20 (51%) of 39 patients with DM had DOF (ie, without LRF), indicating that these patients may have had occult distant disease at the time of diagnosis, and investigating novel diagnostic tools (eg, functional imaging and biomarkers) could identify DM at an earlier phase with the potential of successful salvage therapy.
This study of 300 patients represents the largest published series of OSCC patients treated with postoperative IMRT. Outcomes data were prospectively collected. All patients were uniformly treated with surgery and postoperative IMRT with or without concurrent chemotherapy; however, this may represent selection bias in terms of outcomes. Patients who did not require PORT (in view of favorable disease) or patients who did not undergo initial surgery (in view of inoperable or unresectable disease) were excluded. Furthermore, the number of patients with DOF was too few to allow for multivariable analysis to further identify predictors of DOF. We believe that the pattern of failure in OSCC has shifted, with relatively more frequent DM following postoperative IMRT owing to improved LRC. This is especially important given the lethal impact of DM on OSCC patients. Our results can also help in the identification of a subset of high-risk patients characterized by poor clinical outcomes owing to high metastatic potential with a short latency of metastatic disease. These factors could also be employed for trials of treatment intensification or screening strategies for DM in the early years after surgery.
Surgery and postoperative IMRT with or without concurrent chemotherapy achieved encouraging outcomes with low rates of severe late toxic effects in OSCC. The extent and clinicopathologic characteristics of DOF and DM with locoregional failure were similar. Patients with the combination of G2-3, pN2-3 and ECE have the highest risk of DOF. Both pN2-3 and G2-3 were independent predictors of DM. Further research is required for OSCC patients with these adverse features to determine surveillance and therapeutic strategies targeting DM.
Corresponding Author: Andrew Hope, MD, Princess Margaret Cancer Centre, Department of Radiation Oncology, University of Toronto, 610 University Ave, Rm 5-606, Toronto, ON, M5G 2M9, Canada (email@example.com).
Published Online: December 29, 2016. doi:10.1001/jamaoto.2016.3668
Author Contributions: Dr Hope had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Hosni, Huang, Xu, O'Sullivan, Hope.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hosni, Xu, Bayley, de Almeida, O'Sullivan, Hope.
Critical revision of the manuscript for important intellectual content: Hosni, Huang, Xu, Su, Bayley, Bratman, Cho, Giuliani, Kim, Ringash, Waldron, Spreafico, O'Sullivan, Goldstein, Hope.
Statistical analysis: Hosni, Huang, Xu, Su, O'Sullivan, Hope.
Administrative, technical, or material support: Hosni, Bayley, O'Sullivan, Hope.
Supervision: Xu, Giuliani, O'Sullivan, Hope.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Create a personal account or sign in to: