SEARCH

SEARCH BY CITATION

Keywords:

  • hepatocellular carcinoma;
  • sorafenib;
  • hand-foot skin reaction;
  • vascular endothelial growth factor;
  • tumor necrosis factor-alpha

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

BACKGROUND:

Sorafenib currently sets the new standard for advanced hepatocellular carcinoma (HCC). It has been suggested that Asian patients with HCC have increased susceptibility to hand-foot skin reaction (HFSR) related to sorafenib therapy. The authors investigated the association between sorafenib-induced HFSR and genetic polymorphisms in Korean patients with HCC.

METHODS:

For this prospective cohort study, the authors enrolled 59 consecutive patients with intermediate stage HCC from 5 centers in Korea. All patients received sorafenib 400 mg twice daily in combination with transarterial chemoembolization (TACE). Genotyping was performed on a total of 49 single nucleotide polymorphisms (SNPs) in 8 candidate genes (minor allelic frequency ≥5%). Serum levels of vascular endothelial growth factor (VEGF) and tumor necrosis factor-alpha (TNF-α) were measured using enzyme-linked immunosorbent assays before therapy and 1 month after therapy.

RESULTS:

During a median treatment period of 18 months, 55 patients (93%) developed sorafenib-induced HFSR, including grade 1 reactions in 15 patients, grade 2 reactions in 27 patients, and grade 3 reaction in 13 patients. The SNPs TNF-α −308GG, VEGF −94GG, VEGF 1991CC, VEGF IVS3-28CC, and uridine diphosphate glucuronosyltransferase 1 family-polypeptide A9 (UGT1A9) IVS1-37431AA were associated significantly with the development of high-grade (grade 2 or 3) HFSR in univariate analysis (P < .05). In multivariate analysis, the SNPs VEGF 1991CC (odds ratio, 45.7), TNF-α −308GG (odds ratio, 44.1), and UGT1A9 IVS1-37431AA (odds ratio, 18.7) were identified as independent risk factors for the development of high-grade HFSR (P = .01, P = .02, and P = .02, respectively). He serum TNF-α level measured 1 month after sorafenib therapy was correlated significantly with the development of high-grade HFSR (odds ratio, 3.56; P = .026).

CONCLUSIONS:

Differences in the incidence of HFSR may have been caused by ethnic differences in genetic polymorphisms of the TNF-α, VEGF, and UGT1A9 genes, especially in relation to the expression of serum TNF-α after sorafenib therapy. Cancer 2013. © 2012 American Cancer Society.

Hepatocellular carcinoma (HCC) is the third most common cancer. Approximately 14,000 patients develop HCC annually in Korea,1 and >50% of patients with HCC receive palliative treatments only, including transarterial chemoembolization (TACE), systemic chemotherapy, and radiation therapy.2, 3 Sorafenib, a dual Raf kinase/vascular endothelial growth factor (VEGF) receptor inhibitor, currently sets the new standard for the first-line treatment of advanced HCC.4-6 Recent Asian and Western trials have indicated that Asian patients appear to be more susceptible to hand-foot skin reaction (HFSR) related to sorafenib treatment.4-8 However, little is understood regarding the mechanism underlying HFSR and interracial difference in sorafenib-induced toxicity.

Sorafenib-induced HFSR, which is a common side effect observed in many clinical trials, often is serious and may lead to compromised efficacy because of dose reduction or discontinuation.7-9 Therefore, early detection and management of HFSR is imperative to allow patients to continue life-prolonging therapy with minimal morbidity.

The safety profiles of sorafenib in the Asia-Pacific study5 and in the Sorafenib Hepatocellular Carcinoma Assessment Randomized Protocol (SHARP) study4 were similar. However, there were significant disparities in the incidences of some sorafenib-related adverse events of any grade in both studies. It is noteworthy that the incidence of HFSR was greater in the Asia-Pacific study compared with the SHARP study (45.0% vs 21.2%, respectively), in contrast to the incidence of diarrhea (25.5% vs 39.1%, respectively). The source of these disparities remains unclear but may be caused by physiologic deference associated with ethnicity, including cytochrome P450 3A4 (CYP3A4) and uridine diphosphate glucuronosyltransferase 1A9 (UGT1A9).8, 9 Also, it has been suggested that VEGF and tumor necrosis factor-α (TNF-α) are overexpressed in patients with HCC, and the VEGF pathway is important in the causation of HFSR.8, 10-13 Thus, further investigations on the genetic association between sorafenib-induced HFSR and single nucleotide polymorphisms (SNPs) of VEGF, TNF-α, CYP3A4, and UGT1A9 are warranted to determine the risk of sorafenib-induced HFSR. In this study, we investigated frequencies of SNPs of the genes related to drug metabolism and tumor angiogenesis in ethnically homogenous Korean patients who had HCC with the objective of identifying their probable function as biomarkers of HFSR susceptibility.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

In total, 59 patients with intermediate stage HCC who were enrolled on the Study in Asia of the Combination of TACE with Sorafenib in Patients with Hepatocellular Carcinoma (START) trial from 5 medical centers in Korea were included in the current study (Table 1). The inclusion criteria were an Eastern Cooperative Oncology Group performance status of 0 or 1, a Child-Pugh score ≤7 (Child class A or B), and Barcelona Clinic Liver Cancer (BCLC) stage B disease (intermediate stage; ie, multinodular, asymptomatic tumors without extrahepatic spread). Patients were excluded if they had an infiltrative type of HCC or vascular involvement. Patients also were excluded if they had hypertension, serious nonhealing wounds, and serious acute or chronic illness.

Table 1. Patients and Tumor Characteristics at Baseline, n = 59
VariableMedian [Range]/ No. of Patients (%)
  • Abbreviations: AFP, α-fetoprotein; ALT, alanine aminotransferase; HBV, hepatitis B virus; HCV, hepatitis C virus; INR, international normalized ratio; NBNC, non-HBV/non-HCV hepatitis.

  • a

    Tumor size was based on the greatest dimension of the target lesions.

  • b

    Adapted from: Ueno S, Tanabe G, Nuruki K, et al. Prognostic performance of the new classification of primary liver cancer of Japan (4th edition) for patients with hepatocellular carcinoma: a validation analysis. Hepatol Res. 2002;24:395-403.14

Age, y57 [37-75]
Men52 (88)
White blood count, cells/mm32600 [1150-8824]
Platelets, ×103/mm3126 [66-358]
ALT, IU/L42 [8-294]
Bilirubin, mg/dL0.9 [0.4-2.5]
Albumin, g/dL3.8 [2.8-4.7]
Platelets, INR1.1 [0.95-1.34]
Tumor size, mma48 [20-146]
Serum AFP, ng/mL61 [2-176,000]
Child-Pugh class 
A55 (93)
B4 (7)
Underlying disease 
HBV52 (88)
HCV1 (2)
NBNC6 (10)
TNM stageb 
II25 (42)
III34 (58)

All patients received sorafenib 400 mg twice daily in combination with TACE. Repeated TACE was performed at intervals of 6 to 8 weeks, and sorafenib was administered continuously with dose interruptions 4 days before and after TACE. Treatment-related toxicity was graded using the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 3.0. Tumor response was assessed according to the modified Response Evaluation Criteria in Solid Tumors. We collected peripheral blood samples from each patient 1 week before therapy to obtain serum and genomic DNA samples. The study protocol was approved by the Institutional Review Board of Asan Medical Center, and written informed consent was obtained from all patients who provided blood and tissue samples.

Genotyping of Single Nucleotide Polymorphisms

Peripheral blood samples were drawn into serum separator tubes, centrifuged at ×1800g for 10 minutes, and then stored at −80°C. Genomic DNA was prepared from stored buffy coat that had been extracted from peripheral blood samples using a nucleic acid isolation device (the QuickGene-mini80; Fuji Film, Tokyo, Japan).

Genotyping was performed on a total of 49 SNPs with minor-allele frequency >5% in 8 candidate genes that are involved in angiogenesis and sorafenib metabolism (UGT1A9, UGT1A1, CYP3A4, CYP2B6, TNF-α, VEGF, insulin-like growth factor-2 [IGF2], and hypoxia-inducible factor-1α [HIF-1α]) using the GenomeLab SNPstream genotyping platform (Beckman Coulter, Inc., Fullerton, Calif) and its accompanying SNPstream software suite. The 49 primer pairs used in the multiplex polymerase chain reaction and the single-base extension primers were designed for the specific SNPs and flanking sequences in a homogenous reaction using Web-based software (provided at: http://www.autoprimer.com/; September 2009; Beckman Coulter, Inc.). The 49 individual SNPs were identified by their position and fluorescent color in each well according to the position of the tagged oligonucleotides. Genotype data were generated on the basis of the relative fluorescent intensities for each SNP. Graphic review and operator adjustment of the genotype clusters were performed to refine fluorescent cutoff values.

Serum VEGF and TNF-α Measurement

Serum VEGF and TNF-α concentrations were measured quantitatively using an enzyme-linked immunosorbent assay (ELISA) kit (Quantikine Human Immunoassay; R&D Systems, Minneapolis, Minn) according to the manufacturer's instructions.

Statistical Analysis

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

The genotype frequencies were checked by consistency among normal controls with those expected from Hardy-Weinberg equilibrium. Genotype and allele frequencies were compared between groups using the chi-square test or the Fisher exact probability test, as appropriate. Gene frequencies were determined by gene counting. A P value < .05 was considered statistically significant. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated if the chi-square test or Fisher exact test was significant. For multivariate analysis, binary logistic regression analysis was performed to determine the most discriminating factor for the development of high-grade (grade 2 or 3) HFSR. P values were corrected by multiplying by the number of possible genotypes of single SNPs (Bonferroni correction). All statistical tests were performed with the SPSS statistical software package (version 12.0; SPSS Inc., Chicago, Ill).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

Tumor Characteristics

This prospective cohort study included 59 consecutive patients who had BCLC stage B HCC without portal vein invasion. The median tumor size was 4.8 cm (range, 2.0-14.6 cm); and, according to the American Joint Committee on Cancer TNM classification, 42% of patients had stage II HCC, and 58% of patients had stage III HCC (Table 1). The median treatment duration was 18 months (range, 2-24 months). The 1-year and 2-year overall survival rates were 93% and 81%, respectively; and the 1-year and 2-year progression-free survival rates were 66% and 32%, respectively.

Overall Incidence of Drug-Related Adverse Events

During the treatment period, the most frequent adverse events were HFSR (93%), alopecia (69%), skin rash (32%), diarrhea (15%), and hypertension (12%). There were no grade 4 drug-related adverse events in any categories (Table 2). Fifty-five of 59 patients (93%) developed sorafenib-induced HFSR, including grade 1 reactions in 15 patients, grade 2 reactions in 27, and grade 3 reactions in 13. In addition, 47 of 59 patients (80%) had dose reductions, and 3 patients (5%) discontinued sorafenib because of drug-related adverse events. The median time to HFSR occurrence from the start of sorafenib therapy was 2 weeks (range, 1-12 weeks). There were no differences in age, sex, etiology, Child-Pugh class, tumor size, tumor number, or α-fetoprotein (AFP) level between patients with and without high-grade (grade 2 or 3) HFSR after sorafenib therapy (P > .05) (Table 3). However, the frequency of high-grade HFSR was significantly greater in patients who developed their first HFSR within 3 weeks of sorafenib therapy (OR, 1.4; 95% CI, 0.96-1.99; P = .03).

Table 2. Sorafenib-Related Adverse Events, n = 59
Adverse EventNo. of Patients (%)
  1. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; HFSR, hand-foot skin reaction; ULN, upper limit of normal.

HFSR55 (93)
Alopecia41 (70)
Rash19 (32)
Diarrhea9 (15)
Itching8 (14)
Hypertension7 (12)
Fatigue7 (12)
Neutropenia: <1000 neutrophils/mm36 (10)
Low platelets, <50,000 platelets/mm35 (9)
AST/ALT elevation >5×ULN3 (5)
Anorexia3 (5)
Hemorrhage3 (5)
Table 3. Baseline Patient Characteristics Associated With High-Grade (Grade 2 or 3) Hand-Foot Skin Reaction as Determined by Univariate Analysis, n = 59
VariableP
  1. Abbreviations: AFP, α-fetoprotein; ALT, alanine aminotransferase; HFSR, hand-foot skin reaction.

Age: ≤50 y vs >50 y.517
Sex: Men vs women.827
Child-Pugh class: A vs B.058
AFP: ≤200 ng/mL vs >200 ng/mL.060
ALT: ≤50 IU/L vs >50 IU/L.361
Platelets: ≤100x103/mm3 vs >100x103/mm3.598
Tumor size: ≤5 cm vs >5 cm.698
No. of tumors: ≤2 vs >2.277
TNM stage: II vs III.977

Genotype Frequencies Associated With the Risk of High-Grade Hand-Foot Skin Reaction

Among the studied SNPs, TNF-α −308GG, VEGF −94GG, VEGF 1991CC, VEGF IVS3-28CC, and UGT1A9 IVS1-37431AA were associated significantly with developing high-grade HFSR in univariate analysis (P < .05) (Table 4). In addition, multivariate analysis using a logistic regression model indicated that VEGF 1991CC (OR, 45.7), TNF-α −308GG (OR, 44.1), and UGT1A9 IVS1-37431AA (OR, 18.7) were independent risk factors for developing high-grade HFSR (P = .01, P = .02 and P = .02, respectively) (Table 5).

Table 4. Genotypes of Selected Polymorphisms Associated With Developing High-Grade (Grade 2 or 3) Hand-Foot Skin Reaction as Determined by Univariate Analysis, n = 59
   No. of Patients (%)  
SNPMajor/Minor AlleleGenotypePatients With Grade 2 or 3 HFSRPatients Without Grade 2 or 3 HFSRORP
  1. Abbreviations: A, adenine; C, cytosine; G, guanine; HFSR, hand-foot skin reaction; OR, odds ratio; SNP, single nucleotide polymorphism; T, thymine; TNF-α, tumor necrosis factor-alpha; UGT1A9, uridine diphosphate glucuronosyltransferase 1 family-polypeptide A9; VEGF, vascular endothelial growth factor.

TNF-α −308G/AGG39 (97.5)14 (73.7)13.93.011
  GA1 (2.5)5 (26.3)  
VEGF −94G/CGG17 (42.5)1 (5.3)13.30.005
  CG, CC23 (57.5)18 (94.7)  
VEGF 1991C/TCC38 (95)15 (78.9)10.13.036
  CT1 (2.5)4 (21.1)  
VEGF IVS3-28C/TCC17 (42.5)2 (10.5)6.28.017
  CT, TT23 (57.5)17 (89.5)  
UGT1A9 IVS1-37431A/GAA27 (67.5)7 (36.8)3.56.026
  GA, GG13 (32.5)12 (63.2)  
Table 5. Independent Risk Factors Associated With High-Grade (Grade 2 or 3) Hand-Foot Skin Reaction as Determined by Multivariate Analysisa
VariableOR95% CIP
  • Abbreviations: A, adenine; AFP, α-fetoprotein; C, cytosine; CI, confidence interval; G, guanine; HFSR, hand-foot skin reaction; OR, odds ratio; TNF-α, tumor necrosis factor-alpha; UGT1A9, uridine diphosphate glucuronosyltransferase 1 family-polypeptide A9; VEGF, vascular endothelial growth factor.

  • a

    Variables that had P values ≤.2 in univariate analyses were subjected to multivariate analysis.

  • b

    These variables had a statistically significant association.

TNF-α −308GG44.061.688-1149.907.023b
VEGF −94GG2.590.258-26.081.418
VEGF 1991CC45.682.412-865.029.011b
VEGF IVS3-28CC2.580.279-23.828.404
UGT1A9 IVS1-37431AA18.7171.762-198.838.015b
First HFSR within 3 wk6.000.775-46.531.086
Child Pugh class A0.040.001-1.388.075
AFP ≤200 ng/mL7.900.942-66.296.057
No. of tumors >20.040.002-0.909.043

Serum VEGF and TNF-α Levels

To analyze the association of serum VEGF and TNF-α levels with the development of high-grade (grade 2 or 3) HFSR, we used log10(VEGF) and log10(TNF-α) levels because of their skewed distribution. There were no differences in serum VEGF or TNF-α levels between patients with and without high-grade HFSR. However, high-grade HFSR developed more frequently in the high serum TNF-α group (log10[TNF-α] level, >1.4; OR, 3.56; 95% CI, 1.158-10.930; sensitivity, 67.5%; specificity, 63.2%; P = .026) compared with the low serum TNF-α group 1 month after the therapy. It is interesting to note that patients who had the TNF-α −308GG genotype also had significantly higher serum TNF-α levels 1 month after therapy compared with patients who had the other genotypes (mean ± standard deviation, 1.50 ± 0.25 pg/mL vs 1.24 ± 0.09 pg/mL; log value; P < .001) (Fig. 1).

thumbnail image

Figure 1. Patients who had the tumor necrosis factor-alpha (TNF-α) −308GG genotype had significantly higher serum levels of TNF-α 1 month after therapy compared with patients who had the other genotypes (mean ± standard deviation [SD], 1.50 ± 0.25 pg/mL vs 1.24 ± 0.09; pg/mL; P < .001; log-rank test).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

Like other antineoplastic agents, sorafenib is associated with many side effects, including diarrhea, nausea, fatigue, hypertension, and dermatologic toxicities. HFSR is currently emerging as a major toxicity of sorafenib treatment that requires clinical management and dose modifications, although the mechanism underlying HFSR is not clearly understood. However, the reported incidence of HFSR varies significantly among different clinical trials, ranging between 9.1% and 61.9%.7, 15-19 Recent Asian and Western trials have indicated that Asian patients have increased susceptibility to HFSR related to sorafenib treatment.5 We also observed that the incidence of sorafenib-induced HFSR in our Korean patients with HCC was higher than that reported in previous trials.5 The relatively high incidence of HFSR in our study may have been a result of the concomitant administration of doxorubicin for TACE. However, in a randomized controlled trial of TACE using doxorubicin and gelatin-sponges, there was no report of HFSR among the treatment-related adverse events.3 Therefore, the higher risk of HFSR occurrence in the current study in relation to sorafenib therapy is much more likely to be associated with genetic characteristics of out patients than the combination of TACE or the use of doxorubicin.

In our series, >85% of patients experienced HFSR within 1 month of sorafenib administration. It is interesting to note that the incidence of high-grade HFSR was significantly greater in patients who experienced their first HFSR within 3 weeks of sorafenib therapy. These findings caution us to maintain a high index of suspicion for the development of high-grade HFSR when the first reaction appears so early.

The pathogenesis of sorafenib-associated HFSR is uncertain. Recently, several hypotheses regarding the pathophysiology of HFSR have been suggested. These include: 1) the accumulation of potentially toxic local concentrations in eccrine sweat glands, which present in greatest number or density in the palms and soles; 2) damaged vascular integrity because of the dual VEGFR-2 and platelet-derived growth factor inhibition by sorafenib; and 3) keratinocyte injury from sorafenib inhibition of tyrosine protein kinase kit (c-kit) or Raf kinase.7, 20 The significant association of VEGF polymorphisms with high-grade HFSR in our study possibly supports the above-mentioned hypotheses in relation to damaged vascular integrity.

TNF-α and VEGF are important mediators of both inflammation and angiogenesis.10 Angiogenesis plays an important role in the inflammatory reaction. Inflammatory infiltrate leads to the development of fibrosis, yielding an increase of resistance to blood flow, which, in turn, leads to tissue hypoxia.21 This condition can stimulate the release of proangiogenetic factors responsible for vascular remodeling and ischemic tissue injury. Damaged vascular integrity and consequent tissue ischemia characterize HFSR.8 In our study, several SNPs relevant to tumor angiogenesis, especially TNF-α and VEGF, increased the risk of sorafenib-induced high-grade HFSR, suggesting that the altered expression of TNF-α and VEGF may mediate the development and the severity of most dermatologic adverse events after sorafenib therapy.

TNF-α is a proinflammatory cytokine and an indirect mediator of angiogenesis.22, 23 A genetic polymorphism of TNF-α at position −308 of the promoter region, which includes the TNF-α −308G and TNF-α −308A alleles, has been associated with susceptibility to various types of cancer.24 SNPs in promoter regions of these gene can influence the expression levels of these mediators.25 Carriers with the TNF-α −308G allele tend to produce higher levels of the cytokine.26 These reports coincide with the higher serum TNF-α level in patients with the TNF-α −308GG genotype observed in our study cohort 1 month after sorafenib therapy. In addition, TNF-α −308GG was identified as an independent risk factor for developing high-grade HFSR. Also, high-grade HFSR developed more frequently in the high serum TNF-α group compared with the low serum TNF-α group 1 month after therapy, but not at baseline. These findings suggest that TNF-α overexpression induced by sorafenib administration may have mediated the genetic predisposition to HFSR after sorafenib therapy in our patients with HCC.

At high levels, TNF-α has antivascular and antiangiogenic activity; whereas, at lower concentrations, TNF-α promotes angiogenesis.27 In addition, TNF-α reduces blood flow in tumors in a dose-dependent fashion.28 These observations indicate that the TNF-α −308GG genotype may contribute to the development of HFSR by increasing circulating TNF-α levels by means of the poor vascular exchange. It is well known that TNF-α is an important mediator of septic shock.29 It is known that natural or recombinant TNF-α has considerable side effects, including fatal vascular collapse and the induction of widespread inflammatory responses.23 This “septic shock-like syndrome” by increased serum TNF-α may be responsible for the causation of HFSR associated with impaired acral tissue circulation.

In the International HapMap cohorts, which included large-scale genotyping in trios, TNF-α −308G allele frequencies were much higher in Chinese populations (allele frequency, 0.967) and in Japanese populations (allele frequency, 0.977) than in Caucasian populations (allele frequency, 0.783).30 VEGF 1991C allele frequencies both in Japanese populations (allele frequency, 0.830) and in our Korean population (allele frequency, 0.941) were a little bit higher than that in Caucasian populations (allele frequency, 0.828).30 Also, UGT1A9 IVS1-37431A frequencies were notably higher in Chinese populations (allele frequency, 0.795) and in Japanese populations (allele frequency, 0.727) than in Caucasian populations (allele frequency, 0.583). Therefore, a possible explanation for the high incidence rate of HFSR among Japanese, Chinese, and Korean populations may be associated with the aforementioned high prevalence of the frequencies of these gene in Asians.

The current study was performed with a relatively small number of patients; thus, it had limited power to establish these genetic associations with clinical events. However, we have demonstrated functional associations between the genes relevant to angiogenesis and their expression in relation to developing high-grade HFSR. These findings may provide some mechanistic explanation for the development of HFSR in relation to the VEGF pathway. In addition, the finding of an association between the TNF-α pathway and the development of HFSR is very interesting pathophysiologically and deserves further study.

In conclusion, we report that several SNPs relevant to tumor angiogenesis or drug metabolism, especially VEGF 1991CC, TNF-α −308GG, and UGT1A9 IVS1-37431AA, increase the risk of sorafenib-induced, high-grade HFSR. These findings suggest that the differences in the reported incidence of HFSR among different clinical trials may be related to ethnic differences in the allele frequencies of the above-described SNPs. Also, the higher serum TNF-α levels observed in our study 1 month after therapy were correlated significantly with developing high-grade HFSR. These data warrant further investigations on VEGF and TNF-α to identify the underlying pathophysiologic mechanism of HFSR.

Note Added in Proof

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Statistical Analysis
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. Note Added in Proof
  9. REFERENCES