Activating mutations in RAS are frequently present in patients with acute myeloid leukemia (AML), but their overall prognostic impact is not clear.
Activating mutations in RAS are frequently present in patients with acute myeloid leukemia (AML), but their overall prognostic impact is not clear.
A retrospective analysis was performed to establish the clinical characteristics of patients with RAS-mutated (RASmut) AML, to analyze their outcome by therapy, and to describe the proteomic profile of RASmut compared with wild-type RAS (RASWT) AML.
Of 609 patients with newly diagnosed AML, 11% had RASmut. Compared with RASWT, patients with RASmut AML were younger (median age, 54 years vs 63 years; P = .001), had a higher white blood cell count (16K mm−3 vs 4K mm−3 ; P < 0.001) and bone marrow blast percentage (56% vs 42%; P = .01) at diagnosis, and were less likely to have an antecedent hematologic disorder (36% vs 50%; P = .03). The inv(16) karyotype was overrepresented in patients with RASmut and the −5 and/or −7 karyotype was underrepresented. RAS mutations were found to have no prognostic impact on overall survival or disease-free survival overall or within cytogenetic subgroups. There was a suggestion that patients with RASmut benefited from cytarabine (AraC)-based therapy. Proteomic analysis revealed simultaneous upregulation of the RAS-Raf-MAP kinase and phosphoinositide 3-kinase (PI3K) signaling pathways in patients with RASmut.
RAS mutations in AML may delineate a subset of patients who benefit from AraC-based therapy and who may be amenable to treatment with inhibitors of RAS and PI3K signaling pathways. Cancer 2012. © 2012 American Cancer Society.
Activating mutations in the RAS protooncogene are among the most common genetic alterations in human cancer.1–3 They are present overall in approximately 40% of all malignancies, and noted in up to 90% of pancreatic adenocarcinoma, 30% of non-small cell lung cancer, and approximately one-third of colorectal cancers.1, 2 Several series have reported RAS mutations in 10% to 25% of cases of acute myeloid leukemia (AML)4–14 and 7% to 48% of cases of myelodysplastic syndrome.15–17 N-RAS, K-RAS, and H-RAS are the most common isoforms of RAS involved in human cancer. The activating point mutations most frequently occur in codons 12, 13, or 61, although less common point mutations also have been described.2, 18 RAS is a guanosine-5′-triphosphate (GTP)-dependent second messenger protein that couples signals from receptor tyrosine kinases, cytokine receptors, and growth factor receptors to intracellular signaling networks to induce proliferation and survival (Fig. 1).2, 3, 19, 20 Under normal conditions, an upstream activation signal promotes the binding of GTP to RAS, facilitated by GTP exchange factors. This activation leads to the phosphorylation of downstream Raf, Mek, and the mitogen-activated protein (MAP) kinase-extracellular signal-regulated kinase (ERK). ERK can then translocate to the nucleus and modulate transcription.2, 20 This process is self-regulated by the GTPase activity associated with RAS, which hydrolyzes the GTP and turns the signal off. Oncogenic mutations of RAS that disrupt this GTPase activity lead to aberrant constitutive signaling that is independent of the upstream stimuli.2, 3, 19, 20 There is also important crosstalk between the RAS-Mek-MAP kinase pathway and other growth and survival pathways such as the phosphoinositide 3-kinase (PI3K)-Akt pathway (Fig. 1). In malignancies in which the RAS pathway is constitutively active due to mutations, it is important to recognize that additional associated pathways such as the PI3K-Akt axis may provide redundant growth and survival signals. An important mechanism of resistance to the therapeutic targeting of RAS is crosstalk and signaling through alternate pathways.20 Determining the activation status of proteins downstream from RAS and PI3K in samples from patients with AML with mutated RAS (RASmut) versus wild-type RAS (RASWT) could potentially yield additional therapeutic targets and mechanisms of resistance.
The impact of RAS mutations on outcome and response to therapy in various cancers is currently under extensive investigation.11, 21, 22 Recent work has shown that RAS mutations in patients with AML may be predictive of a benefit from postremission therapy with high-dose cytarabine (AraC).11 However, the overall prognostic impact of RAS mutations in patients with AML is not clear. Although some investigators reported that RAS mutations were associated with an adverse outcome,9, 14, 17 others have suggested a favorable impact,7, 10 and several have noted no significant prognostic impact.4, 6, 12, 13 In the current study, we attempted to further characterize the role of RAS mutations in patients with AML. We analyzed the baseline characteristics of patients with newly diagnosed AML with RASmut or RASWT and evaluated their response to different therapies and their overall outcome. We also examined their pretreatment proteomic profiles to determine any differences in the 2 cohorts.
The records of patients with newly diagnosed AML who were seen at The University of Texas MD Anderson Cancer Center between 2003 and 2007 were analyzed. This included 609 patients in whom RAS mutational analysis was performed at the time of diagnosis. A total of 228 patients, who were representative of the entire group, had banked material for proteomic analysis. All patients underwent a bone marrow examination at diagnosis. Informed consent was appropriately obtained from all patients, according to the institutional guidelines and in compliance with the Declaration of Helsinki (local protcols: PA11-0788, Lab 01- 473, Lab 05-0654) . AML was diagnosed, according to the World Health Organization criteria,23 by the presence of at least 20% myeloid blasts in the blood and/or bone marrow.
Bone marrow cytogenetics were performed by conventional chromosomal banding and/or fluorescence in situ hybridization when needed. The cytogenetics were classified into 3 prognostic subgroups: favorable (t(8;21), inv(16), t(16;16), or t(15;17) ), adverse (−5, −7, 11q23 abnormality, or complex karyotype [≥3 cytogenetic abnormalities] ), or intermediate (neither adverse nor favorable).
RAS mutation analysis was performed on unsorted cells from diagnostic bone marrow aspirate samples (with a median blast percentage of 44) for exon 1 (codons 12 and 13) and exon 2 (codon 61) for both N-RAS and K-RAS using polymerase chain reaction (PCR)-based DNA sequencing methods. FLT3(fms-related tyrosine kinase 3) mutation analysis was performed on the same samples using PCR-based DNA sequencing methods.
Proteomic analysis of matched, banked pretreatment patient blood or bone marrow samples was performed using an automated, high-throughput reverse phase protein array as previously described.24, 25 Briefly, protein lysates from patient samples were printed onto methylcellulose-coated slides in 5 serial dilutions along with normalization and expression controls. Each slide was probed with a distinct validated antihuman antibody for the protein or phosphoprotein in question and a secondary antibody to amplify the signal. Finally, a stable dye was precipitated.26 The stained slides were analyzed using MicroVigene software (Vigene Tech, Carlisle, Mass) to produce quantified data. A total of 176 validated antibodies were used. Supercurve algorithms were used to generate a single value from the 5 serial dilutions.27 Loading control28 and topographical normalization procedures accounted for protein concentration and background staining variations. Associations between protein expression levels and RAS mutation status were assessed in R using standard Student t tests.
Patients received several different treatment regimens depending on the available clinical trials as well as their age and performance status. For the purposes of this analysis, induction therapy could be divided into 3 groups: 1) a high-dose AraC-based regimen (≥ 1000 mg/m2/day [HiAraC]), 2) a conventional dose AraC-based regimen (100 mg/m2/day to 500 mg/m2/day [ConvAraC]), and 3) a non-AraC–based regimen (non-AraC). Treatments within the non-AraC cohort included hypomethylating agents (azacitidine, decitabine), other nucleoside analogues such as clofarabine, and small-molecule targeted therapies, among others (Table 1).
|Category||Treatment||No. of Patients (%)|
|High-dose AraC based||Ida + HiDAC||209 (34)|
|Flud + HiDAC||60 (10)|
|Flud + HiDAC + Ida||1 (0)|
|Dauno + HiDAC||2 (0)|
|Misc HiDAC||4 (1)|
|Clofarabine + HiDAC||38 (6)|
|Conventional-dose AraC based||Misc ConvAraC||101 (17)|
|NonAraC based||Clofarabine||34 (6)|
|Arsenic+ATRA (APL)||11 (2)|
|Misc No AraC, Other||13 (2)|
Complete remission (CR) was defined by the presence of < 5% blasts in the bone marrow, the absence of extramedullary leukemia, and peripheral blood count recovery with a neutrophil count of at least 1 × 109/L and a platelet count of at least 100 × 109/L.29 Treatment failure was defined by the absence of a documented CR after therapy, including induction death or refractory disease. Disease recurrence was defined by an excess of 10% leukemic blasts in a bone marrow aspirate that was not related to normal hematopoietic recovery or the development of new extramedullary leukemia. Disease-free survival was calculated from the time of first documented CR to disease recurrence or death in CR. The duration of CR (CRd) was calculated from the time of first documented CR to disease recurrence. Overall survival (OS) was calculated from the time of diagnosis of AML to death from any cause.
Patient and disease characteristics were compared using chi-square tests. These characteristics are listed in Table 2 and include age, sex, white blood cell (WBC) count, platelet count, beta2-microglobulin level, blast percentage, French-American-British (FAB) morphology, cytogenetic category, FLT3- internal tandem duplication (ITD) positivity, and history of antecedent hematologic disorder (AHD). Estimates of survival curves and CRd were calculated according to the Kaplan-Meier product limit method and compared by means of the log-rank test. The Cox proportional hazards regression model was used to assess the relation between patient characteristics and survival. Predictive variables with P < .1 by univariate analysis were included in a multivariate analysis. All reported P values were 2-sided, and those < .05 were considered statistically significant. A multivariate logistic regression analysis was conducted assessing complete response to treatment as a dichotomous variable (response vs no response). Independent predictors included in the model were adverse cytogenetics, AHD, RAS mutation, performance status, use of AraC therapy, and diagnosis of therapy-related AML. Two-way interactions between the variables in the model were explored and reported and the model goodness of fit was verified using the Hosmer-Lemeshow chi-square test. We used a P < .05 as the level of statistical significance. The statistical analyses were accomplished using IBM SPSS statistical software for Windows (version 19.0; SPSS Inc, Chicago, Ill).
|Characteristic||Mutated RAS N=66 (11% )||Wild-Type RAS N= 543 (89%)||P|
|Median age (range), y||54 (25-88)||63 (17-88)||.001|
|Male sex, no. (%)||34 (52)||284 (52)||.9|
|Median WBC count (range), × 103/mm3||16 (0.8-111.7)||4.2 (0.3-433)||<.001|
|Median platelets (range), × 103/mm3||43 (7-357)||50 (2-676)||.14|
|Median PB blast (range), %||22 (0-95)||10 (0-97)||.007|
|Median BM blast (range), %||56 (20-93)||42 (0-98)||.014|
|FAB morphology, no. (%)|
|M0||0 (0)||23 (4)||<.001|
|M1||4 (6)||55 (10)|
|M2||15 (23)||113 (21)|
|M3||0 (0)||11 (2)|
|M4||25 (38)||80 (15)|
|M5||11 (17)||36 (7)|
|M6||0 (0)||29 (5)|
|M7||0 (0)||1 (0)|
|Cytogenetics, no. (%)|
|inv(16)||10 (15)||16 (3)||<.001|
|t(8;21)||3 (5)||16 (3)|
|t(15;17)||0 (0)||11 (2)|
|-5 and/or -7||8 (12)||132 (24)|
|11q abnormality||3 (5)||16 (3)|
|Other complex karyotypea||6 (9)||50 (9)|
|Diploid, -Y||27 (41)||233 (43)|
|Other||9 (14)||67 (13)|
|History of AHD, no. (%)||24 (36)||274 (50)||.03|
|FLT3 ITD-positive, no. (%)||3 (5)||64 (12)||.08|
|Median β2-microglobulin||3.6 (1.2-12.9)||2.7 (1.1 - 20.8)||.002|
From January 2003 to December 2007, 609 patients with newly diagnosed AML were identified in whom RAS mutational analysis was performed. Of these, 66 patients (11%) were found to have RAS mutations: 48 (73%) with an N-RAS mutation, 16 (24%) with a K-RAS mutation, and 2 patients with both. Patient characteristics of the RASmut cohort and the RASWT cohort are summarized in Table 2. The median age for patients with RASmut was 54 years versus 63 years for those with RASWT (P = .001). In addition, patients with RASmut had a significantly higher WBC count (median, 16K/mm3 vs 4K/mm3; P < .001), peripheral blast percentage (median, 22% vs 10%; P = .007), bone marrow blast percentage (median, 56% vs 42%; P = .01), and beta2-microglobulin (median, 3.6 vs 2.7; P = .002) at diagnosis. Patients with RASmut were less likely than those in the wild-type cohort to have a history of AHD (36% vs 50%; P = .03). A higher percentage of patients in the wild-type cohort had FLT3-ITD, but this did not reach statistical significance. The prevalence of AML with RASmut among patients with different FAB morphologic subtypes and different karyotypes is depicted in Figure 2. There was a significantly higher prevalence of RASmut among patients with FAB M4 (acute myleomonocytic leukemia; 24% [P < .001]) and M5 (acute monocytic leukemia; 23% [P = .004]) histologies. RAS mutations were significantly overrepresented in inv(16), having been found in 39% of patients with this karyotype (P < .001). Conversely, RAS mutations were significantly underrepresented in the −5/-7 and t(15;17) karyotypes (P = .03).
This cohort of patients with newly diagnosed AML was treated with a variety of different therapies according to their age, performance status, karyotype, and the availability of specific clinical trials (Table 1). Because these are among the covariates determined to be important in predicting outcome in patients with AML, we first performed a multivariate analysis of predictive factors for the achievement of CR (Table 3). Independent predictors of the achievement of CR in the entire cohort were age, cytogenetics, history of AHD, and treatment with AraC-based therapy. Notably, the presence of a RAS mutation and performance status were not found to be significant. We then analyzed the OS and CRd for the entire cohort of patients by RAS mutation status, regardless of treatment or cytogenetics. There was no significant difference in OS or CRd between patients with AML with RASmut or RASWT (Fig. 3). Among patients with the inv(16) karyotype, there was no significant difference noted in OS or CRd between patients with RASmut or RASWT (Fig. 4 Top). Among patients with an intermediate-risk karyotype, there was no significant difference in OS or CRd between those patients with RASmut or RASWT (Fig. 4 Middle). In patients with an adverse risk karyotype, there was a trend toward worse OS and CRd in patients with RASmut, but this did not reach statistical significance (Fig. 4 Bottom). Among patients with RASmut, there was no significant difference in OS between those who had an N-RAS versus K-RAS mutation (median, 12.9 months vs 14.5 months; P = .78) (Fig. 5). Similarly, there was no significant difference in the CR rate between patients with N-RAS or K-RAS (56% vs 63%; P = .66)
|Age (continuous variable)||<.001|
|History of AHD (yes/no)||<.001|
|Treatment (AraC vs no AraC)||<.001|
|Prior chemo or XRT (yes/no)||NS|
|RAS mutation (present/absent)||NS|
|ECOG PS (continuous)||NS|
The response rates by treatment in the 66 patients with RASmut are outlined in Table 4 and compared with patients with RASWT. There was a significantly higher rate of CR in patients who received any AraC versus those who received no AraC (70%, 70%, and 0%, respectively, for HiAraC, ConvAraC, and nonAraC; P < .001). Because clinical factors such as age and history of AHD may influence the decision to treat patients with AraC-based or non–AraC-based therapy, we next examined the outcome of patients with RASmut by age (< 60 years vs ≥ 60 years), a diagnosis of secondary AML (vs primary or de novo AML), and whether they had received AraC-based therapy.(Fig. 6) All of the 39 patients aged < 60 years in this cohort received AraC-based therapy and had the best outcome (median OS, 23.5 months). Among the older patients with RASmut, those who received AraC had a significantly improved OS compared with those who did not (median OS, 12 months vs 2.5 months). When the patients were analyzed by their diagnosis of primary versus secondary AML, those patients who had primary AML and received AraC-based therapy had the longest OS (median, 26.5 months), compared with those who had either secondary AML (treated with or without AraC) or those who had primary AML and were treated with non–AraC-based therapy. Indeed, on multivariate analysis, the factors that were found to independently predict OS in patients with RASmut were adverse cytogenetics (P = .016) and a history of AHD (P = .001). (Table 5).
|Treatment||Mutated RAS (N=66)||Wild-Type RAS (N=543)|
|No. (%)||CRa (%)||No . (%)||CR (%)|
|High-dose AraC||44 (67)||31 (70)||269 (50)||167 (62)|
|Standard-dose AraC||10 (15)||7 (70)||91 (17)||56 (62)|
|No AraC||12 (18)||0 (0)||182 (34)||63 (35)|
|Age (continuous variable)||.095|
|History of AHD||.001|
|Treatment (AraC vs no AraC)||NS|
|Prior chemo or XRT||NS|
|WBC count (continuous)||.07|
|BM blast, % (continuous)||.08|
Of the 609 cases analyzed, 228 patients (37%) had banked material available for proteomic analysis. The baseline characteristics of these patients were representative of the entire cohort (Table 6). High-throughput analysis of these 228 samples was performed using a panel of 176 validated antibodies. Of the 176 proteins tested, 39 (22%) were differentially expressed between baseline AML samples with RASWT and RASmut (Table 7). With a false discovery rate of 5%, the expectation would be to find 8.8 proteins being significantly different, and therefore this frequency was significantly higher than expected. Twenty proteins were expressed at relatively higher levels in samples with RASmut and 19 were relatively higher in samples with RASWT (Table 7). Among the proteins that were expressed at relatively higher levels in samples with RASmut were a substantial number of downstream targets and effectors of both the RAS-Raf-MAP kinase pathway as well as the PI3K-Akt pathway such as total (P = .003) and phospho-MEK (P = .04), total (P = .004) and phospho-p38 (P = .1), phospho-Akt (P = .01), phospho-glycogen synthase kinase (GSK) (P = .002), phospho-ERK (P = .1), and total (P = .048) and phospho–phosphoinositide-dependent kinase-1 (PDK1)(P = .009) (Table 7).
|Characteristic||Mutated RAS N=30 (13% )||Wild-Type RAS N= 198 (87%)|
|Median age (range), y||57 (28-84)||69 (17-87)|
|Male sex, no. (%)||16 (53)||114 (58)|
|Median WBC count, × 103/mm3||17||6|
|Median PB blast, %||29||13|
|FAB morphology, no. (%)|
|M0||0 (0)||8 (4)|
|M1||3 (10)||25 (13)|
|M2||9 (30)||79 (40)|
|M3||0 (0)||4 (2)|
|M4||8 (27)||34 (17)|
|M5||5 (17)||18 (9)|
|M6||3 (10)||14 (7)|
|M7||0 (0)||1 (1)|
|Cytogenetics, no. (%)|
|Favorable||7 (23)||15 (8)|
|Intermediate||13 (43)||91 (46)|
|Adverse||10 (33)||92 (47)|
|History of AHD, no. (%)||7 (23)||82 (41)|
|FLT3 ITD-positive, no. (%)||3 (10)||36 (18)|
|Proteins Higher in Mutated RAS||Proteins Higher in Wild-Type RAS|
To further characterize the significance of RASmut in patients with newly diagnosed AML, we performed a retrospective, single-institution, clinical and proteomic analysis on a large cohort of patients who were treated heterogeneously. We aimed to describe the clinical characteristics of these patients, to analyze the effect of different treatments on outcome, to assess the prognostic impact of RASmut in AML, and to examine the differential protein expression in the leukemic cells of patients with RASmut and RASWT. We found several pretreatment clinical characteristics that were significantly different between patients with RASmut and RASWT. Patients with RASmut in the study cohort tended to be younger than those with RASWT and less likely to have AHD. Similar to other reported series,9, 13 patients with RASmut had a higher median WBC count at the time of presentation. However, in contrast with other reports,10, 11, 13 the current study indicated that patients with RASmut had higher peripheral blood and bone marrow blast percentages at presentation.
Confirming observations from other large studies of patients with AML, RAS mutations were significantly overrepresented in patients with the inv(16) karyotype and underrepresented in patients with t(15;17) and −5/-7 karyotypes.4, 6 This strong association with inv(16) suggests a cooperative interaction between the 2 genetic abnormalities. Activation of the RAS pathway in patients with inv(16) may be an important mechanism of pathogenesis and should also be investigated further in patients with inv(16) without RAS mutations for other routes of pathway activation. If this signaling pathway is indeed mechanistically important in patients with inv(16) AML, then targeting this pathway could offer therapeutic opportunities. Conversely, one explanation for the relative underrepresentation of RAS mutations in patients with t(15;17) could be the presence of alternate genetic abnormalities that can activate constitutive signaling through the RAS-MAPK and related pathways. For example, FLT3-ITD mutations, which are present in approximately 40% of cases30–32 with t(15;17), are upstream of and exploit the RAS-MAPK and PI3K-Akt signaling pathways. Indeed, we and others6, 9, 13 have shown that RAS mutations and FLT-ITD rarely coexist in the same patient with AML. This suggests deregulation of common final signaling pathways resulting from different aberrantly activated upstream molecules.19 If this is confirmed, specifically inhibiting a limited number of signaling pathways could be broadly applicable in patients with AML.
In the current study, RASmut had no clear prognostic significance in AML overall or within the different cytogenetic subgroups in terms of OS or duration of remission. This is in keeping with the majority of other large series.4, 6, 12 Because the cohort in the current study had been treated with a variety of different therapies, we were able to assess the effect of various doses of AraC and nonAraC on outcome in patients with RASmut. Recently, investigators from the Cancer and Leukemia Group B reported that patients with AML with RASmut derived greater benefit from postremission therapy with high-dose AraC compared with those with RASWT and those who did not receive HiDAC consolidation.11 These clinical data confirmed previous in vitro work by Koo et al33, 34 that demonstrated that RASmut AML cell lines were more sensitive to the cytotoxic effects of AraC than those with RASWT. The investigators demonstrated that cells harboring the RAS mutation had a compromised S-phase checkpoint in response to chemotherapy-induced DNA damage.33 Unlike cells with RASWT, which arrested in S-phase in response to AraC, cells with RASmut progressed through the S-phase, incorporating AraC and continuing on to apoptosis.33 This led to a cytotoxic response versus a cytostatic response to AraC. Clinical results from the current study support both the preclinical and previous clinical observations highlighting the sensitivity of RASmut AML to AraC.8, 11, 33, 34
All patients in the current study who received any AraC as part of their induction therapy received HiDAC as their consolidation therapy. Patients with RASmut who were not treated with AraC had a lower CR rate and shorter OS than those who received AraC as part of their induction therapy. This difference in OS was also observed in the subgroup of older patients with RASmut, and, although not statistically significant, in those with AHD and RASmut. It is interesting to note that patients with RASmut and a history of AHD had the poorest OS, regardless of their therapy. On multivariate analysis, the only factors that were found to be associated with OS in patients with RASmut were a history of AHD and cytogenetics. This may suggest a degree of selection bias in the current study cohort, because older patients or those with a predetermined adverse prognosis who were not able to tolerate intense chemotherapy were offered clinical trials not containing AraC, thereby accounting for the diminished effect of treatment on the overall prognosis noted on multivariate analysis.
Exploring the differential protein expression signature of patients with RASmut AML may give provide insight into the drivers of this disease and potentially highlight new therapeutic targets in this patient subset. Using a high-throughput proteomics array, we were able to compare the pretreatment protein expression signature from samples of patients with AML and RASmut compared with those with RASWT. We found a significant number of proteins that were expressed differentially between AML samples containing RASmut and RASWT. As hypothesized, several downstream targets and effectors of the RAS-Raf-MAP kinase pathway as well as the PI3K-Akt pathway were expressed at relatively higher levels in samples with RASmut. Although this needs to be validated further, this finding suggests a greater activation of both signaling pathways in patients with RASmut AML, compared with RASWT AML, with the implication that they may respond to small molecule inhibitors of kinases along these pathways. Illmer et al8 conducted a similar investigation on the activation status of the RAS pathway in patients with AML. Using Western blot analysis to detect activated RAS (as defined by Raf-bound RAS), the investigators found an activated RAS pathway in approximately 25% of all AML cases. It is interesting to note that only 22% of the RAS-mutated cases were found to have increased RAS activity using their method. However, the study did not simultaneously examine the activation of associated signaling pathways such as the PI3K-Akt pathway, which may have been relatively upregulated in the AML samples with RASmut. To our knowledge, the current study is the first to simultaneously assess the activation status of protein signaling networks in patients with AML and RASmut.
In conclusion, we found that AML with RASmut has distinct clinical features and an association with certain cytogenetic and morphologic subgroups. Although we did not find a prognostic effect of RAS mutations overall in patients with AML, we did find an interaction between AraC-based treatment or a history of AHD and overall outcome. Comparative proteomic profiling of RASmut AML suggests upregulation and activation of proteins involved in the RAS-Raf-MAPK and PI3K-Akt signaling networks that are relatively higher than in AML without RASmut. This may form the basis of future studies investigating the use of specific inhibitors and their combinations in this subset of patients.
We thank Mark Brandt and Marylou Cardenas for assistance with statistical analysis and the generation of survival curves.
This study was supported in part by National Cancer Institute grant K12CA088084 (to T.M.K.).
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.