A total of 6900 patients were identified with clinical stage III to IVA oral cavity squamous cell carcinoma treated with surgery and postoperative radiotherapy or definitive chemoradiotherapy.
Kaplan-Meier analysis of survival among all patients and in a propensity score–matched cohort of patients treated with S+PORT or CRT (A and B); among all patients and in a propensity score–matched cohort of patients treated with surgery and postoperative radiotherapy alone (S+RT), surgery plus postoperative chemoradiotherapy (S+CRT), or CRT (C and D); among all patients treated with CRT or S+PORT grouped by academic or nonacademic institutions (E and F); and in propensity score–matched cohorts of patients with T1 to T2 or T3 to T4a OCSCC (G and H). The log-rank test was used to assess for differences in overall survival. Δ indicates difference.
eTable. Patient Characteristics in the Propensity Score–Matched Population
Customize your JAMA Network experience by selecting one or more topics from the list below.
Spiotto MT, Jefferson G, Wenig B, Markiewicz M, Weichselbaum RR, Koshy M. Differences in Survival With Surgery and Postoperative Radiotherapy Compared With Definitive Chemoradiotherapy for Oral Cavity CancerA National Cancer Database Analysis. JAMA Otolaryngol Head Neck Surg. 2017;143(7):691–699. doi:10.1001/jamaoto.2017.0012
Do surgery and postoperative radiotherapy improve survival compared with definitive chemoradiotherapy for oral cavity squamous cell cancer?
In an analysis of 6900 patients with stage III to IVA oral cavity squamous cell cancer identified in the National Cancer Database, surgery and postoperative radiotherapy were associated with improved survival compared with chemoradiotherapy in the entire population and in a propensity score–matched cohort.
For operable locally advanced oral cavity squamous cell cancer, surgery as the initial treatment modality may improve survival.
Because locally advanced oral cavity squamous cell carcinoma (OCSCC) is often treated with surgery followed by postoperative radiotherapy (S+PORT), the effectiveness of organ preservation with concurrent chemoradiotherapy (CRT) remains unclear.
To compare the differences in survival between patients with locally advanced OCSCC treated with S+PORT or CRT.
Design, Setting, and Participants
Using the National Cancer Database, this study compared 6900 patients with stage III to IVA OCSCC treated with S+PORT and CRT from 2004 through 2012 at academic and community-based cancer clinics. Comparisons were made using Kaplan-Meier methods and Cox proportional hazards regression models using the entire cohort and a propensity score–matched cohort of 2286 patients.
Main Outcomes and Measures
Overall survival (OS).
Of the 6900 study patients, 4809 received S+PORT (3080 male [64.0%] and 1792 [36.0%] female) and 2091 received CRT (1453 male [69.5%] and 638 [30.5%] female). Median follow-up for the entire group was 23.0 months overall but was shorter for patients receiving CRT (17.3-month) vs S+PORT (25.6 months). Patients receiving CRT were more likely to be older than 60 years, treated before 2007, live within 10 miles of the treating facility, treated at nonacademic centers, have more comorbidities, have T3 to T4a tumors, and have N2a to N2c nodal disease. Propensity score matching identified cohorts of patients with similar clinical variables. S+PORT was associated with improved survival among all patients (3-year OS: 53.9% for S+PORT vs 37.8% for CRT; difference = 16.1%; 95% CI, 13.6%-18.6%) and in the propensity score–matched cohort (3-year OS: 51.8% for S+PORT vs 39.3% for CRT; difference = 11.9%; 95% CI, 7.8%-16.0%). S+PORT was associated with improved survival among patients with T3 to T4a tumors (3-year OS: 49.7% for S+PORT vs 36.0% for CRT; difference = 16.1%; 95% CI, 13.6%-18.6%) but was not associated with improved survival among patients with T1 to T2 tumors (3-year OS: 59.1% for S+PORT vs 53.5% for CRT; difference = 5.6%; 95% CI, −3.1% to 14.3%).
Conclusions and Relevance
Compared with CRT, S+PORT was associated with improved survival for locally advanced OCSCCs, especially in T3 to T4a disease. These data support the use of surgery as the initial treatment modality for operable OCSCCs.
In locally advanced oral cavity squamous cell carcinomas (OCSCCs), surgery followed by postoperative radiotherapy with or without chemotherapy (S+PORT) has been the preferred treatment because of improved reconstruction techniques and the poor outcomes of locally advanced OCSCCs treated with definitive radiotherapy.1-5Quiz Ref ID Nevertheless, the morbidities associated surgical resection followed by radiotherapy or chemoradiotherapy (CRT) include worse dysphagia and increased use of feeding tubes.6 Although organ preservation with CRT has similar outcomes compared with S+PORT in other cancers of the pharyngeal axis,7-9 studies10,11 have found conflicting results regarding the equivalence of CRT and S+PORT for OCSCC. Pederson et al10 reported 20-year locoregional control rates approaching 90% and overall survival (OS) rates of 76%. By contrast, other studies1-5,12 have reported substantially worse outcomes for definitive CRT in locally advanced OCSCC. Scher et al4 reported 2-year locoregional control rates of 41% with concurrent CRT. Sher et al5 reported 2-year local regional control rates of 64% for CRT compared with 91% for S+PORT. Furthermore, in a randomized clinical trial of S+PORT compared with CRT, the oral cavity subset of patients undergoing CRT for head and neck squamous cell carcinomas fared significantly worse than those receiving S+PORT.3
However, groups reporting poor outcomes with definitive CRT in patients with OCSCC rely on sample sizes of 12 to 73 patients, which makes it difficult to compare outcomes to S+PORT series, in which often substantially greater numbers of patients are analyzed. Furthermore, the inferior outcomes with CRT have resulted from inclusion of patients with worse performance status or with unresectable cancers. Because of these limited numbers of patients treated with CRT, it is difficult to compare outcomes of S+PORT and CRT in patients who are matched for confounding clinical variables known to affect disease control and/or survival. Consequently, the limited numbers and variability of patients in previous studies2,4,5,10 make it difficult to draw conclusions regarding the efficacy of CRT compared with S+PORT. This study assesses the survival of 6900 patients with locally advanced OCSCC treated with S+PORT or curative CRT in the entire patient population and in a propensity score–matched cohort.
Quiz Ref IDThis study used the National Cancer Database (NCDB), which is a hospital-based registry and is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. The CoC's NCDB and the hospitals participating in the CoC NCDB are the source of the deidentified data used in this study; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by us. The NCDB has set criteria to ensure that the data submission by each cancer center meets prespecified quality benchmarks. The University of Illinois at Chicago Institutional Review Board waived approval.
This analysis included patients 18 years or older who had stage III to IVA head and neck squamous cell cancers that involved the oral cavity who received at least part of the initial treatment at the reporting facility and definitive S+PORT or concurrent CRT as part of their treatment. The University of Illinois at Chicago Institutional Review Board waived patient consent. Patients were excluded if they had evidence of in situ disease, clinical stage I to II or IVB disease, or distant metastatic disease or were treated with palliative intent. Patients were treated from 2004 through 2012 because 2004 was the first year the NCDB began collecting detailed information on radiation treatment and 2012 was the last year in the database with survival information. The cohort composition as derived from the whole database is shown in Figure 1.
Demographic variables included age, sex, comorbidity, year of diagnosis, distance from treating facility, and facility type. Age was grouped into 4 categories, 50 years or younger, 51 through 60 years, 61 through 70 years, and older than 70 years, because these increasing deciles of age were associated with worse survival as estimated by Cox proportional hazards regression modeling. Comorbidity index was based on the Charlson/Deyo comorbidity score: 0 comorbid conditions, 1 comorbid condition, or 2 or more comorbid conditions. Year of diagnosis was grouped into 3 categories: 2004 through 2006, 2007 through 2009, and 2010 through 2012. Distance from treating facility was calculated based on the zip codes of the patients’ residences and treating facilities and grouped into 5 categories: less than 10 miles, 10 through 19 miles, 20 through 21 miles, 30 miles or more, or unknown. Facility type was grouped into 4 categories: academic or research program, comprehensive community cancer program, community cancer program or integrated network cancer program, and unknown. Clinical variables included stage grouping (cSG I-II vs cSG III-IV), clinical and pathologic tumor stage (cT1-2 vs cT3-4a and pT1-2 vs pT3-4a), and clinical and pathologic N stage (cN0-1 vs cN2a-2c and pN0-1 vs pN2a-3). Chemotherapy was defined as single or multiagent chemotherapy given during radiotherapy.
Data were analyzed using SPSS statistical software, version 23 (IBM Inc) and JMP, version 9 (SAS Institute Inc). All tests were 2-sided. The χ2 test was used to compare between discrete variables and the t test between continuous variables. Differences between medians were assessed using the Wilcoxon test. Differences and 95% CIs were calculated. Multivariable stepwise logistic regression analyses were performed to calculate independent predictors for the receipt of definitive concurrent CRT. Survival curves were plotted based on the Kaplan-Meier method. Cox proportional hazards regression was used to determine the effect of surgery and adjuvant radiotherapy or concurrent CRT on survival while adjusting for known confounding variables using patient and treatment characteristics that were different between cohorts. Hazard ratios (HRs) and 95% CIs were estimated in models that adjusted for the covariates of interest. Survival analysis was performed for all patients and for a subset of propensity score–matched patients. Propensity score matching was performed using SPSS to create balanced cohorts of patients treated with S+PORT or CRT. The SPSS estimates the propensity score by using logistic regression in which S+PORT or CRT was used as the outcome variable and the pretest covariates as the predictors. Patient and treatment variables (age, sex, year of diagnosis, distance from treating facility, facility type, comorbidity index, cT stage, cN stage, and tumor site) were selected as pretest covariates because these variables had demonstrated differences between treatment groups and/or have been associated with differences in survival for patients with head and neck cancer. The propensity score was used to generate 1:1 nearest neighbor–matched cohorts with a caliper width (maximum allowable difference between 2 participants) of 0.0001 of the SD of the logit propensity score. The adequacy of the match was assessed using standardized mean differences (Cohen d). The 2 cohorts were compared using the log-rank test, with the HR for survival calculated using univariate Cox proportional hazards regression.
Of the 6900 study patients, 4809 received S+PORT (3080 male [64.0%] and 1792 [36.0%] female) and 2091 received CRT (1453 male [69.5%] and 638 [30.5%] female). Median follow-up for the entire group was 23.0 months overall but was shorter for patients receiving CRT (17.3 months) vs S+PORT (25.6 months) (Table 1). Between groups, patients had similar comorbidity scores and age. Patients treated with CRT were more likely to be male (69.5% vs 64.0%; difference = 5.5%; 95% CI, 3.1%-7.9%), treated in earlier periods (2004-2006: 28.8% vs 19.2%; difference = 9.6%; 95% CI, 7.4%-11.8%), treated closer to home (<10 miles: 54.6% vs 38.0%; difference = 16.6%; 95% CI, 14.1%-19.1%), and treated at comprehensive community cancer centers (41.9% vs 24.9%; difference = 17.0%; 95% CI, 14.6%-19.4%). In addition, patients treated with CRT were more likely to have cT3 to cT4a tumors (75.6% vs 67.5%; difference = 8.1%; 95% CI, 5.8%-10.4%), cN2a to cN2c nodal staging (51.6% vs 40.0%; difference = 11.6%; 95% CI, 9.0%-14.2%), and primary tumors that involved the oral tongue (47.0% vs 35.7%; difference = 11.3%; 95% CI, 8.8%-13.8%).
Compared with S+PORT, receipt of CRT was associated with age of 11 years or older than 70 years (odds ratio [OR], 1.38; 95% CI, 1.15-1.68), male sex (OR, 1.31; 95% CI, 1.16-1.49), earlier period of treatment (2004-2006: reference; 2007-2009: OR, 0.79; 95% CI, 0.68-0.91; 2010-2012: OR, 0.48; 95% CI, 0.42-0.46), less than 10 miles from the treatment facility (<10 miles: reference; 10-19 miles: OR, 0.63; 95% CI, 0.54-0.74; 20-29 miles: OR, 0.67; 95% CI, 0.55-0.81; >30 miles: OR, 0.49; 95% CI, 0.42-0.57), treatment at a comprehensive community cancer center (OR, 2.65; 95% CI, 2.33-3.01), treatment at a community cancer center (OR, 2.54; 95% CI, 2.15-3.01), and comorbidity scores greater than 1 (comorbidity index 1: reference; comorbidity index 2: OR, 0.60; 95% CI, 0.51-0.71; comorbidity index 3: OR, 0.81; 95% CI 0.61-1.06) (Table 2). In addition, receipt of CRT was associated with cT3 to cT4a tumors (OR, 2.11; 95% CI, 1.85-2.41), cN2a to cN2c tumors (OR, 1.81; 95% CI, 1.62-2.04), and oral tongue subsites (OR, 1.78; 95% CI, 1.58-2.01).
In the entire cohort, the 3-year OS for all patients was 53.9% for those treated with S+PORT compared with 37.8% for those treated with CRT (difference = 16.1%; 95% CI, 13.6%-18.6%) (Figure 2A). A total of 2291 patients survived 3 years in the entire cohort. On univariate analysis, improved survival was also associated with younger age, treatment at an academic medical center, more recent eras of treatment, distance to the treating facility, lower comorbidity scores, cT1 to cT2 disease, and cN0 to cN1 disease. Sex and tumor site were not associated with improved survival. Quiz Ref IDOn multivariate analysis, improved survival was associated with S+PORT (HR, 0.66; 95% CI, 0.61-0.71) (Table 3), whereas worse survival was associated with increasing age, treatment at a comprehensive community cancer center, treatment between 2004 and 2006, higher comorbidity index, cT3 to cT4a tumors, and cN2a to cN2c nodal disease.
We generated a propensity score–matched cohort for patients treated with S+PORT and CRT by matching patients based on significantly different patient and tumor variables in Table 1. Propensity score–matched patients treated with CRT or S+PORT did not differ by sex, year of diagnosis, distance, facility type, cT stage, cN stage, and oral cavity subsite (eTable in the Supplement). Prematched standardized mean differences were as follows: age, −0.06 (95% CI, −0.09 to −0.03); sex, 0.12 (95% CI, 0.10-0.13); year of diagnosis, 0.31 (95% CI, 0.29-0.33); facility type, −0.35 (95% CI, −0.37 to 0.33); distance from treating facility, 0.33 (95% CI, 0.30-0.37); comorbidity index, 0.13 (95% CI, 0.12-0.14); cT stage, −0.19 (95% CI, −0.20 to −0.18); cN stage, −0.22 (95% CI, −0.24 to −0.21); and site, 0.24 (95% CI, 0.22-0.25). Postmatched standardized mean differences were as follows: age, 0.0 (95% CI, −0.04 to 0.04); sex, 0.0 (95% CI, −0.19 to 0.19); year of diagnosis, 0.0 (95% CI, −0.04 to 0.03); facility type, 0.0 (95% CI, −0.04 to 0.03); distance from treating facility, 0.0 (95% CI, −0.05 to 0.05); comorbidity index, 0.0 (95% CI, −0.01 to 0.01); cT stage, 0.0 (95% CI, −0.02 to 0.02); cN stage, 0.0 (95% CI, −0.02 to 0.02); and site, 0.0 (95% CI −0.02 to 0.02). In the propensity score–matched cohort, the 3-year OS for all patients was 51.8% for those treated with S+PORT compared with 39.9% for those treated with CRT (difference = 11.9%; 95% CI, 7.8%-16.0%) (Figure 2A and B). Because approximately half of the patients in the S+PORT group received postoperative radiotherapy (S+RT) or postoperative chemoradiotherapy (S+CRT), we assessed the association of S+RT, S+CRT, or CRT on survival in the entire patient population and in the propensity score–matched cohort (Figure 2C and D). In the entire patient population, S+RT was associated with improved survival compared with S+CRT (HR, 1.11; 95% CI, 1.02-1.21) or CRT (HR, 1.74; 95% CI, 1.61-1.88), However, in the propensity score–matched patient population, S+RT was not associated with improved survival compared with S+CRT (HR, 1.06; 95% CI, 0.91-1.26) but was associated with improved survival compared with the CRT (HR, 1.50; 95% CI, 1.30-1.72).
We also assessed the extent to which treatment with S+PORT or CRT was associated with improved survival based on the regionalization of cancer care. When patients were stratified by treatment type, patients treated with S+PORT had improved survival when treated at academic centers (HR, 0.86; 95% CI, 0.78-0.95) compared with comprehensive community cancer centers on univariate analysis. By contrast, patients treated with CRT had similar survival when treated at academic or comprehensive community cancer centers. However, no difference was found in 3-year OS when comparing academic centers with comprehensive community cancer centers when grouped with other nonacademic centers (Figure 2E and F).
The benefit of S+PORT was also observed in patients with advanced tumor stage because patients with cT3 to cT4 disease had a 3-year OS of 49.7% when treated with S+PORT compared with 36.0% when treated with CRT (difference = 13.7%; 95% CI, 9.1%-18.3%) (Figure 2G and H). By contrast, in patients with cT1 to cT2 disease, the 3-year OS was not significantly different (59.1% for S+PORT vs 53.5% for CRT; difference = 5.6%; 95% CI, −3.1% to 14.3%). S+PORT was associated with improved survival among patients with cN0 to cN1 nodal disease and cN2a to cN2c nodal disease. On multivariate analysis, S+PORT, younger age, and cT1 to cT2 and cN0 to cN1 disease remained independently significant for improved OS.
This study found that patients with OCSCC treated with S+PORT had improved OS compared with patients treated with CRT. Quiz Ref IDThe differences in survival benefit were most apparent for patients with advanced tumor stage because S+PORT benefited patients with cT3 to cT4a tumors but not patients with cT1 to cT2 tumors. Compared with CRT, the benefit of S+PORT was unlikely attributable to improved survival among patients receiving S+CRT because there was no difference in survival between S+RT and S+CRT in the propensity score–matched analysis. The likelihood of receiving CRT was associated with older age, earlier era of treatment, advanced tumor and nodal stage, and oral tongue primary site. Although CRT was more likely associated with more advanced tumor and nodal stages, our analysis included only patients with stage III to IVA disease to exclude potentially unresectable N3 and/or T4b disease that potentially would negatively bias survival outcomes with CRT. We further minimized bias in patient selection between the S+PORT and CRT cohorts by generating a propensity score–matched cohort to account for differences in patient comorbidity, treatment era, tumor stage, and other significantly different variables that may confound outcomes. Quiz Ref IDConsequently, our data strongly suggest that initial surgery is associated with improved survival compared with CRT for patients with locally advanced OCSCC.
The OS rates attained with S+PORT in this study are consistent with previously reported survival outcomes for adjuvant radiotherapy in OCSCC. Our survival rates are similar to the OS rate reported by Daly et al13 for adjuvant radiotherapy in 30 patients with OCSCC, of whom 44% of patients had T3 to T4 disease and 76% had stage III to IV disease. By contrast, Gomez et al14 and Yao et al15 reported slightly higher 2-year OS rates of 68% to 74% for adjuvant radiotherapy for OCSCC, with 40% to 56% of patients having T3 to T4a disease and 80% to 91% of patients having stage III to IV disease. Sher et al5 reported even higher 2-year OS rates for adjuvant radiotherapy of 85%; however, only 26% of patients had T3 to T4a disease, and 64% of patients had stage III to IV disease. Therefore, S+PORT outcomes in a national database are similar to other single institutional series for adjuvant radiotherapy when accounting for tumor stage.
However, the OS obtained with CRT was significantly worse than the survival for S+PORT. These findings are consistent with the long-term results of a randomized clinical trial that compared S+PORT with CRT for resectable head and neck squamous cell carcinoma.3,16 Although that trial did not find a survival benefit with S+PORT, a subgroup analysis of patients with OCSCC suggested that S+PORT improved 5-year disease-specific survival compared with CRT (68% for S+PORT vs 12% for CRT). However, that trial suffered from poor accrual because it enrolled 119 of a planned 200 patients, of whom only 32 patients had OCSCC. Similarly, retrospective reviews2,4,5 of patients treated with CRT demonstrated relatively worse survival outcomes for CRT compared with S+PORT. Sher et al5 reported 2-year OS rates of 63% among 12 patients treated with definitive CRT, which was significantly worse than the OS of 85% in a separate cohort treated with S+PORT. In 50 patients treated with CRT, Gore et al2 reported worse OS among patients treated with CRT compared with S+PORT. However, these conclusions are limited by differences in patient cohorts because the CRT group had 2.3-fold more patients with T4 disease and 1.8-fold more patients with stage IV disease than the S+PORT group. A retrospective series that evaluated CRT for patients with inoperable OCSCC also reported poor rates of OS (5-year OS of 15%).4 Although the poor results with CRT were associated with cisplatin-based regimens, CRT regimens using hyperfractionated radiotherapy (fluorouracil and hydroxyurea) have been reported to produce high rates of locoregional control approaching 90% and OS of 66.9% at 5 years, findings similar to those in a subset of patients treated with surgery alone.10,11 It remains unclear why fluorouracil and hydroxyurea–based CRT regimens were associated with better outcomes compared with cisplatin-based CRT regimens, especially given that a phase 2 trial found no difference in locoregional control or OS between cisplatin-based and fluorouracil and hydroxyurea–based CRT regimens for head and neck squamous cell carcinomas.17 Consequently, our analysis is consistent with other series that suggest that initial surgery likely improves OS over CRT for patients with locally advanced OSCC.1-5
Several studies18-20 have examined the effect of regionalization of treatment on cancer care. Bhattacharyya and Abemayor19 found that head and neck oncologic care is increasingly being regionalized to academic medical centers. Furthermore, compared with nonteaching institutions, a higher proportion of major ablative procedures for patients with head and neck cancer are being performed at academic institutions, which also has been associated with improved survival among patients with OCSCC.18,20 Similarly, compared with comprehensive community cancer centers, we found that academic centers treated more patients with OCSCC with S+PORT, which was associated with improved survival at these academic centers. Because survival outcomes were similar between academic and nonacademic centers for patients treated with CRT, it is tempting to speculate that improved survival among patients treated with S+PORT at academic centers may, in part, be attributable to increased surgical expertise associated with higher patient volumes at these academic centers.
As with any nonrandomized study, our study has several limitations. Because the NCDB does not include patterns of local, regional, or distant recurrence, we can only assume that differences in OS were in part attributable to disease recurrence. Consequently, we cannot exclude survival differences attributable to differences in patient and tumor variables not included in the database. In addition, a substantial number of patients were excluded because of incomplete or inaccurate information, noncurative treatment intent, and disease that was in situ, early stage, or unresectable. Our analysis also lacked specific information regarding chemotherapy agents used, quality-of-life measures, and adverse events for each modality. Furthermore, we cannot exclude that patients treated with CRT were inoperable because of comorbidities and/or unresectable disease, which was not captured by the tumor stage, nodal stage, and/or the Charlson/Deyo comorbidity index. In addition, this information was collected across multiple institutions; consequently, we cannot account for center-specific differences in treatment protocols and treatment techniques. On the other hand, our results are not limited by particular treatment paradigms of a single institution. Therefore, despite the limitations of our study, our data from a large cohort analysis indicate that S+PORT results in better outcomes than CRT for locally advanced OCSCC, consistent with several single-institutional series.
This study found that S+PORT improves survival compared with CRT for locally advanced OCSCC. Given the improved results with S+PORT, this study further supports the premise that organ preservation should not be the primary goal for treating locally advanced OCSCC with CRT.
Corresponding Author: Michael T. Spiotto, MD, PhD, Department of Radiation and Cellular Oncology, University of Chicago, Knapp Center for Biomedical Discovery 6142, 900 E 57th St, Chicago, IL 60637 (firstname.lastname@example.org).
Accepted for Publication: December 28, 2016.
Published Online: April 20, 2017. doi:10.1001/jamaoto.2017.0012
Author Contributions: Dr Spiotto had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Spiotto, Wenig, Markiewicz, Koshy.
Acquisition, analysis, or interpretation of data: Spiotto, Jefferson, Weichselbaum, Koshy.
Drafting of the manuscript: Spiotto, Markiewicz, Koshy.
Critical revision of the manuscript for important intellectual content: Spiotto, Jefferson, Wenig, Weichselbaum, Koshy.
Statistical analysis: Spiotto, Koshy.
Obtained funding: Spiotto.
Administrative, technical, or material support: Wenig, Markiewicz, Weichselbaum, Koshy.
Study supervision: Weichselbaum.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: Dr Spiotto is supported by grant 1010964 from the Burroughs Wellcome Career Award for Medical Scientists.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.