Abstract
Background: A recent study of familial and early onset prostate cancer reported a recurrent rare germline mutation of HOXB13 among men of European descent. The gene resides within the 17q21 hereditary prostate cancer linkage interval.
Methods: We evaluated the G84E germline mutation (rs138213197) of HOXB13 in a case–control study of familial prostate cancer at Vanderbilt University (Nashville, TN) to independently evaluate the association of the mutation with familial prostate cancer. We genotyped 928 familial prostate cancer probands and 930 control probands without a personal or family history of prostate cancer.
Results: Our study confirmed the association between the G84E mutation of HOXB13 and risk of prostate cancer among subjects of European descent. We observed the mutation in 16 familial cases and in two controls, each as heterozygotes. The odds ratio (OR) for prostate cancer was 7.9 [95% confidence interval, (CI) 1.8–34.5, P = 0.0062] among carriers of the mutation. The carrier rate was 1.9% among all familial case probands and 2.7% among probands of pedigrees with ≥3 affected. In a separate case series of 268 probands of European descent with no additional family history of prostate cancer, the carrier rate was 1.5%.
Conclusions: The germline mutation G84E of HOXB13 is a rare but recurrent mutation associated with elevated risk of prostate cancer in men of European descent, with an effect size that is greater than observed for previously validated risk variants of genome wide association studies.
Impact: This study independently confirms the association of a germline HOXB13 mutation with familial prostate cancer. Cancer Epidemiol Biomarkers Prev; 21(8); 1348–53. ©2012 AACR.
Introduction
Prostate cancer is estimated to have the largest heritable risk component of all common cancers, roughly twice that of breast cancer (1, 2). Family history remains the best clinical predictor of risk. Segregation analyses have been most consistent with a rare genetic component of age-dependent penetrance. The collective results of linkage studies of familial prostate cancer suggest complex heritability: incomplete penetrance (mutations associated with more modest effect sizes than typical of simple Mendelian disease), polygenic inheritance (multiple loci acting jointly to cause disease), and genetic heterogeneity (underlying causal mutations in many genes, each infrequent). These obstacles have posed a marked challenge for the discovery of gene mutations underlying familial prostate cancer. Genome-wide association studies (GWAS) of prostate cancer have detected validated risk variants, but these have been of low effect size (ORs, 1.1–2.0) and collectively have accounted for approximately 25% of observed heritable risk (3–19). These do not explain the observed inheritance patterns of hereditary prostate cancer pedigrees (20). The power of a GWAS to detect uncommon genetic variants of large effect might be anticipated to be low, given that only 3% to 5% of prevalent cases meet criterion for hereditary prostate cancer and given that commercial chips used were designed to detect common disease variants.
With this background in mind, the recently described germline mutation G84E of HOXB13 among familial prostate cancer cases is a noteworthy discovery (21). This mutation resides within a replicated linkage interval (22–25) on 17q21-22 and was detected by sequencing genes of this interval among probands of linked hereditary prostate cancer pedigrees. The germline mutation was observed in 4 of 85 (4.7%) linked pedigrees of European descent and cosegregated with disease. The 85 linked pedigrees were a subset of those used for genetic mapping, most without evidence of linkage to the region. Within a case–control series of that study, the mutation was detected in 45 of 2,064 (2.2%) familial cases, in 19 of 2,410 (0.8%) sporadic cases, and in one of 1,401 (0.07%) controls. Our study sought independent confirmation of the association of the HOXB13 germline mutation with prostate cancer in familial prostate cancer case probands and controls with no personal or family history of prostate cancer.
Materials and Methods
Study population
We initiated the Familial Prostate Cancer Study (FPCS) as an observational hospital-based study of familial prostate cancer at Vanderbilt University (Nashville, TN) in 2003. The FPCS is among the largest case–control studies of familial prostate cancer (Table 1). We recruited incident familial prostate cancer case probands (≥2 affected first- or second-degree relatives in the pedigree) at the time of treatment for the principal diagnosis of prostate cancer (confirmed by review of pathology), and control probands at the time of routine preventative screening for prostate cancer. Controls had no personal or family history of prostate cancer (first- and second-degree relatives), no known prostate-specific antigen (PSA) > 4 ng/dL or abnormal digital rectal exam, and no history of prostate biopsy. This familial study design provides improved power to detect genetic risk variants (26). Approximately 95% of eligible subjects agreed to participate, with written informed consent under Institutional Review Board (IRB) governance. We matched cases to controls in a 1:1 ratio by race and age, within 2.5 years of age at diagnosis or screen. A small excess of familial cases remained unmatched to controls in the ongoing study. All subjects completed a structured questionnaire of family cancer history, demographics, and grandparental ancestry. The date, age, and PSA level at diagnosis of adenocarcinoma or at screen was recorded for each subject. The initial clinical staging (clinical tumor–node–metastasis, TNM), biopsy Gleason score, treatment modality, and date(s) were recorded. More than 97% of case subjects underwent radical prostatectomy, providing definitive histopathologic diagnoses. For these, surgical pathology staging (pathologic TNM), seminal vesicle invasion, margin status, left and right lobe Gleason scores and sum, and capsular penetration were recorded. Among familial cases of European descent with available pathologic staging, 175 of 844 (21%) were of pT3, pT4, N1, or M1 stage. Pathologic staging was not available from prostatectomy for the remaining 21 (2%) of the 865 familial cases of European descent.
Familial Prostate Cancer Study
| . | . | Controls . | Cases . |
|---|---|---|---|
| Total, no. | 930 | 928 | |
| Caucasian, no. | 830 | 865 | |
| African-American, no. | 100 | 63 | |
| Mean agea, y | 60.1 | 60.1 | |
| Mean no. of brothers | 1.8 | 1.7 | |
| Median PSAa | 0.9 | 5.5 | |
| Age diagnosis ≤ 60, no. | — | 470 | |
| Age diagnosis ≥ 61, no. | — | 458 | |
| Gleason sum ≤ 6, no. | — | 469 | |
| Gleason sum ≥ 7, no. | — | 427 | |
| Affected in pedigree, no.b | 0 | 930 | — |
| 2 | — | 575 | |
| ≥3 | — | 353 |
| . | . | Controls . | Cases . |
|---|---|---|---|
| Total, no. | 930 | 928 | |
| Caucasian, no. | 830 | 865 | |
| African-American, no. | 100 | 63 | |
| Mean agea, y | 60.1 | 60.1 | |
| Mean no. of brothers | 1.8 | 1.7 | |
| Median PSAa | 0.9 | 5.5 | |
| Age diagnosis ≤ 60, no. | — | 470 | |
| Age diagnosis ≥ 61, no. | — | 458 | |
| Gleason sum ≤ 6, no. | — | 469 | |
| Gleason sum ≥ 7, no. | — | 427 | |
| Affected in pedigree, no.b | 0 | 930 | — |
| 2 | — | 575 | |
| ≥3 | — | 353 |
aAt diagnosis for cases, at screen for controls.
bProband plus first- and second-degree affected relatives.
In addition, a series of independent singleton cases, without a family history of prostate cancer among first- or second-degree relatives, was also accrued. This included 271 cases of European descent (mean age of diagnosis 58.4) and 104 of African descent (mean age of diagnosis 61.1); 44% and 27% were diagnosed ≤age 55, respectively. Among the singleton cases of European descent, 68 of 256 (27%) were of pT3, pT4, N1, or M1 stage. Pathologic staging was not available from prostatectomy for the remaining 15 of the 271 total singleton cases.
Genotyping
DNA was extracted from whole blood on an Autopure LS robot using the Puregene DNA Purification System Standard Protocol (Qiagen). DNA was quantified using the PicoGreen dsDNA Quantitation Kit (Invitrogen), imaged with a Molecular Devices/LJL Analyst HT (Molecular Devices). Single-nucleotide polymorphism (SNP) genotyping for rs138213197 was conducted using the TaqMan platform (Applied Biosystems). Mutations were confirmed by Sanger sequencing of identified carriers. Genotypes were successfully obtained for 99.2% of subjects. Genotype data were in Hardy–Weinberg equilibrium among controls, familial cases, and singleton cases of European descent. The SNP was monomorphic without mutation carriers among African-Americans.
Statistical analyses
Unconditional logistic regression analysis was used to estimate prostate cancer OR, adjusted for age. Analyses were restricted to subjects of European descent. ORs, 95% CI, and P values were derived under a dominant model. Hardy–Weinberg equilibrium analysis was conducted using Haploview. An association between genotype and prostate cancer was considered nominally significant if the associated 2-sided P value was less than 0.05. The study sought to replicate a previously observed association. Comparisons of carriers and noncarriers of the A rs138213197 variant were made with respect to age at diagnosis, Gleason score, and advanced stage (tumor stage ≥3, or positive lymph nodes, or metastatic disease). These analyses were conducted using the Wilcoxon rank-sum test for age, the score test for trend for Gleason score, and Fisher exact test for advanced stage.
Results
The G84E germline mutation of HOXB13 (allele A on coding strand, encoding glutamate) was observed in 2 study controls and in 20 cases, each in the heterozygous state. All carriers were of European descent. One control carrier was age 80, with 2 unaffected male siblings and a screening PSA of 1.64. The other control carrier was age 56 with 3 unaffected male siblings and screening PSA level of 0.78. Both men had PSA levels near or below the median level for Caucasians of the respective age groups (27, 28). A low rate of misclassification of controls is expected, because a given control may later develop the disease, and because not all prostate cancer is accompanied by an elevated PSA level.
Characteristics of individual mutation carriers are given in Table 2, and Table 3 presents the OR and significance of prostate cancer association with the G84E germline mutation. Cancers other than of the prostate were common in the pedigrees of these probands, of unknown significance. We compared the familial case study population to controls, and the singleton study population to the same controls. We also evaluated strata of familial cases defined by the number affected in the pedigree, by age of diagnosis, by Gleason score, and by extraprostatic disease at diagnosis (Table 4).
Characteristics of HOXB13 G84E (glutamate) carriers of European descent
| Subject . | Status . | Age diagnosis/screen . | PSAa . | Pedigree typeb . | Stage . | Gleason . | # Cases in pedigreec . | Other reported cancers in pedigreec . | ||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Case | 49 | 4.6 | MM | pT2c | pN0 | pMX | 5 | 4 | Five additonal third-degree relatives with prostate cancer |
| 2 | Case | 69 | 4.8 | MM | pT2c | pN0 | pMX | 7 | 4 | Lung, female breast, hepatocellular, colon |
| 3 | Case | 48 | 4.1 | MM | pT2c | pN0 | pMX | 6 | 3 | Stomach |
| 4 | Case | 58 | 4.0 | MM | pT2c | pN0 | pMX | 6 | 3 | Lung, ovary, female breast |
| 5 | Case | 65 | 6.8 | MM | pT2c | pN0 | pMX | 6 | 3 | Lung, head and neck |
| 6 | Case | 59 | 10.8 | Bilineal | pT2c | pN1 | pMX | 7 | 3 | None |
| 7 | Case | 64 | 6.2 | MM | pT2c | pN0 | pMX | 7 | 3 | Female breast, additional unknown type |
| 8 | Case | 55 | 3.9 | Bilineal | pT3a | pN0 | pMX | 5 | 3 | Leukemia, melanoma, pancreas, ovary, multiple myeloma |
| 9 | Case | 66 | 4.2 | MM | pT3a | pN0 | pMX | 7 | 3 | Lung, ovary |
| 10 | Case | 66 | 6.3 | MM | pT2a | pN0 | pMX | 7 | 2 | Head and neck, eye |
| 11 | Case | 56 | 3.9 | MM | pT2c | pN0 | pMX | 6 | 2 | Melanoma, colon, glioma, Hodgkin's lymphoma, lung |
| 12 | Case | 54 | 4.1 | MM | pT2c | pN0 | pMX | 7 | 2 | Unknown type |
| 13 | Case | 67 | 3.4 | MM | pT2c | pNX | pMX | 7 | 2 | Melanoma, leukemia |
| 14 | Case | 61 | 5.2 | NMM | pT2c | pN0 | pMX | 6 | 2 | Uterine |
| 15 | Case | 64 | >50 | NMM | T2c | NX | MX | 8 | 2 | None |
| 16 | Case | 59 | 15 | NMM | pT3b | pNX | pMX | 9 | 2 | Lymphoma |
| 17 | Case | 41 | 10.3 | Singleton | pT2c | pNX | pMX | 7 | 1 | Ovary, lung |
| 18 | Case | 68 | 8.5 | Singleton | pT2c | pN0 | pM0 | 5 | 1 | None |
| 19 | Case | 50 | 8.1 | Singleton | pT2c | pN0 | pMX | 7 | 1 | Unknown type |
| 20 | Case | 48 | 4.5 | Singleton | pT2c | pN0 | pMX | 7 | 1 | None |
| 21 | Control | 80 | 1.64 | — | — | — | — | — | 0 | None |
| 22 | Control | 56 | 0.78 | — | — | — | — | — | 0 | Cervical, lung |
| Subject . | Status . | Age diagnosis/screen . | PSAa . | Pedigree typeb . | Stage . | Gleason . | # Cases in pedigreec . | Other reported cancers in pedigreec . | ||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Case | 49 | 4.6 | MM | pT2c | pN0 | pMX | 5 | 4 | Five additonal third-degree relatives with prostate cancer |
| 2 | Case | 69 | 4.8 | MM | pT2c | pN0 | pMX | 7 | 4 | Lung, female breast, hepatocellular, colon |
| 3 | Case | 48 | 4.1 | MM | pT2c | pN0 | pMX | 6 | 3 | Stomach |
| 4 | Case | 58 | 4.0 | MM | pT2c | pN0 | pMX | 6 | 3 | Lung, ovary, female breast |
| 5 | Case | 65 | 6.8 | MM | pT2c | pN0 | pMX | 6 | 3 | Lung, head and neck |
| 6 | Case | 59 | 10.8 | Bilineal | pT2c | pN1 | pMX | 7 | 3 | None |
| 7 | Case | 64 | 6.2 | MM | pT2c | pN0 | pMX | 7 | 3 | Female breast, additional unknown type |
| 8 | Case | 55 | 3.9 | Bilineal | pT3a | pN0 | pMX | 5 | 3 | Leukemia, melanoma, pancreas, ovary, multiple myeloma |
| 9 | Case | 66 | 4.2 | MM | pT3a | pN0 | pMX | 7 | 3 | Lung, ovary |
| 10 | Case | 66 | 6.3 | MM | pT2a | pN0 | pMX | 7 | 2 | Head and neck, eye |
| 11 | Case | 56 | 3.9 | MM | pT2c | pN0 | pMX | 6 | 2 | Melanoma, colon, glioma, Hodgkin's lymphoma, lung |
| 12 | Case | 54 | 4.1 | MM | pT2c | pN0 | pMX | 7 | 2 | Unknown type |
| 13 | Case | 67 | 3.4 | MM | pT2c | pNX | pMX | 7 | 2 | Melanoma, leukemia |
| 14 | Case | 61 | 5.2 | NMM | pT2c | pN0 | pMX | 6 | 2 | Uterine |
| 15 | Case | 64 | >50 | NMM | T2c | NX | MX | 8 | 2 | None |
| 16 | Case | 59 | 15 | NMM | pT3b | pNX | pMX | 9 | 2 | Lymphoma |
| 17 | Case | 41 | 10.3 | Singleton | pT2c | pNX | pMX | 7 | 1 | Ovary, lung |
| 18 | Case | 68 | 8.5 | Singleton | pT2c | pN0 | pM0 | 5 | 1 | None |
| 19 | Case | 50 | 8.1 | Singleton | pT2c | pN0 | pMX | 7 | 1 | Unknown type |
| 20 | Case | 48 | 4.5 | Singleton | pT2c | pN0 | pMX | 7 | 1 | None |
| 21 | Control | 80 | 1.64 | — | — | — | — | — | 0 | None |
| 22 | Control | 56 | 0.78 | — | — | — | — | — | 0 | Cervical, lung |
aPSA at diagnosis for case, at screen for control.
bMM designates apparent male-to-male transmission (autosomal dominant), whereas NMM indicates no apparent male-to-male transmission (X-linked or sibship).
cAmong first- and second-degree relatives.
Association of prostate cancer with HOXB13 G84E (rs138213197) on chromosome 17q21.32
| Affected in pedigreea . | Carrier rate (%) . | OR (95% CI) . | P . |
|---|---|---|---|
| 0 | 2 of 825 (0.002) | ||
| 1 | 4 of 268 (0.015) | 5.6 (0.9–33.9) | 0.061 |
| ≥1 | 20 of 1,126 (0.018) | 7.2 (1.7–31.2) | 8.1 × 10−3 |
| 2 | 7 of 529 (0.013) | 5.8 (1.2–28.2) | 0.030 |
| ≥2 | 16 of 858 (0.019) | 7.9 (1.8–34.5) | 6.2 × 10−3 |
| ≥3 | 9 of 329 (0.027) | 11.8 (2.5–55.3) | 1.8 × 10−3 |
| ≥4 | 2 of 125 (0.016) | 7.0 (1.0–51.4) | 0.055 |
| Affected in pedigreea . | Carrier rate (%) . | OR (95% CI) . | P . |
|---|---|---|---|
| 0 | 2 of 825 (0.002) | ||
| 1 | 4 of 268 (0.015) | 5.6 (0.9–33.9) | 0.061 |
| ≥1 | 20 of 1,126 (0.018) | 7.2 (1.7–31.2) | 8.1 × 10−3 |
| 2 | 7 of 529 (0.013) | 5.8 (1.2–28.2) | 0.030 |
| ≥2 | 16 of 858 (0.019) | 7.9 (1.8–34.5) | 6.2 × 10−3 |
| ≥3 | 9 of 329 (0.027) | 11.8 (2.5–55.3) | 1.8 × 10−3 |
| ≥4 | 2 of 125 (0.016) | 7.0 (1.0–51.4) | 0.055 |
aFirst- and second-degree relatives (0, controls; 1, case probands with no others in pedigree affected; 2, case probands with one additional in pedigree affected, etc.).
Association of prostate cancer with HOXB13 G84E in strata of age of diagnosis and disease aggressiveness
| Affected in pedigreea . | Stratum . | Carrier rate (%) . | OR (95% CI) . | P . |
|---|---|---|---|---|
| ≥2 | ≤60b | 8 of 435 (0.018) | 9.9 (1.4–68.6) | 0.020 |
| ≥2 | ≥61 | 8 of 423 (0.019) | 7.7 (1.6–37.0) | 0.011 |
| ≥2 | Gleason ≤ 6 | 7 of 433 (0.016) | 6.4 (1.3–32.1) | 0.023 |
| ≥2 | Gleason ≥ 7 | 9 of 400 (0.023) | 9.9 (2.1–46.0) | 3.5 × 10−3 |
| ≥2 | ≥pT3, N1, or M1 | 4 of 175 (0.023) | 9.7 (1.8–53.5) | 9.2 × 10−3 |
| ≥3 | ≤60 | 5 of 177 (0.028) | 13.9 (1.8–105.5) | 0.011 |
| ≥3 | ≥61 | 4 of 152 (0.026) | 10.2 (1.8–57.8) | 8.9 × 10−3 |
| ≥3 | Gleason ≤ 6 | 5 of 180 (0.028) | 11.3 (2.0–62.4) | 5.5 × 10−3 |
| ≥3 | Gleason ≥ 7 | 4 of 135 (0.030) | 13.1 (2.4–72.4) | 3.2 × 10−3 |
| ≥3 | ≥pT3, N1, or M1 | 3 of 57 (0.053) | 23.3 (3.8–143.0) | 6.8 × 10−4 |
| Affected in pedigreea . | Stratum . | Carrier rate (%) . | OR (95% CI) . | P . |
|---|---|---|---|---|
| ≥2 | ≤60b | 8 of 435 (0.018) | 9.9 (1.4–68.6) | 0.020 |
| ≥2 | ≥61 | 8 of 423 (0.019) | 7.7 (1.6–37.0) | 0.011 |
| ≥2 | Gleason ≤ 6 | 7 of 433 (0.016) | 6.4 (1.3–32.1) | 0.023 |
| ≥2 | Gleason ≥ 7 | 9 of 400 (0.023) | 9.9 (2.1–46.0) | 3.5 × 10−3 |
| ≥2 | ≥pT3, N1, or M1 | 4 of 175 (0.023) | 9.7 (1.8–53.5) | 9.2 × 10−3 |
| ≥3 | ≤60 | 5 of 177 (0.028) | 13.9 (1.8–105.5) | 0.011 |
| ≥3 | ≥61 | 4 of 152 (0.026) | 10.2 (1.8–57.8) | 8.9 × 10−3 |
| ≥3 | Gleason ≤ 6 | 5 of 180 (0.028) | 11.3 (2.0–62.4) | 5.5 × 10−3 |
| ≥3 | Gleason ≥ 7 | 4 of 135 (0.030) | 13.1 (2.4–72.4) | 3.2 × 10−3 |
| ≥3 | ≥pT3, N1, or M1 | 3 of 57 (0.053) | 23.3 (3.8–143.0) | 6.8 × 10−4 |
aFirst- and second-degree relatives (0, controls; 1, case probands with no others in pedigree affected; 2, case probands with one additional in pedigree affected).
bProband age of diagnosis.
The point estimate of the OR, adjusted for age, was 7.9 (P = 6.2 × 10−3; 95% CI, 1.8–34.5) in a comparison of familial cases and controls. The estimate was greater among cases with a family history of three or more affected (OR = 11.8), relative to a family history of only 2 affected (OR = 5.8). The estimate for cases with no additional family history (OR = 5.6) was similar to the latter but insignificant. Effect size was not further increased among cases with a family history of 4 or more affected (OR = 7.0). Given the rarity of the mutation, the CIs of the OR estimates were broad.
The association appeared similar in familial cases above and below the mean age of diagnosis; there was no significant difference in age of diagnosis between mutation carriers and noncarriers. The point estimates of the OR were higher among familial cases with less differentiated or with extra-prostatic disease at diagnosis. However, there was no significant difference between mutation carriers and noncarriers with respect to either index of aggressive disease.
Discussion
Our study confirms the recently described association of the germline G84E mutation with familial prostate cancer. The association was most evident in pedigrees with 3 or more affected men. OR estimates for the mutation were generally lower in our study than originally estimated. Nonetheless, the effect size appears to be much greater than for other validated associations detected in GWAS investigations.
Given the rarity of the mutation, power was limited to assess the relative impact of the mutation upon age of diagnosis or upon disease aggressiveness. Ewing and colleagues observed a mean age at diagnosis of 52.9 for carriers, and 57.1 for noncarriers, a significant difference in their study. The average age of diagnosis of carriers in our familial prostate cancer study was 58.4; this age was nominally but not significantly different than the 60.2 mean age of diagnosis of noncarriers. Ewing and colleagues found no evidence to support a difference between Gleason grade between G84E carriers and noncarriers and did not investigate an effect on advanced stage prostate cancer. Our analysis observed a nominally higher carrier frequency among cases with higher Gleason score and among cases with extra-prostatic disease at diagnosis, but these differences were not statistically significant.
In Cancer Genome Anatomy Project SAGE data, HOXB13 is expressed in human stomach, colon, and particularly prostate, and expression is reduced in cancers of these sites. The homeodomain protein suppresses both TCF4- (Wnt pathway) and androgen receptor–mediated transcriptional activation to inhibit cell proliferation (29–32). A HOXB13 loss-of-function mutation that leads to an increased risk of prostate cancer is consistent with known prostate cancer biology. However, HOXB13 expression is increased in late stage, hormone-refractory prostate cancer and may also participate in progression to androgen-independent cell proliferation (33). An intriguing correlative question is whether mutation carriers may be predisposed to develop prostate cancer but not to progress to androgen independence. Furthermore, HOXB13 expression is regulated by FOXA1, a gene identified as recurrently somatically mutated in prostate cancer through recent exome sequencing efforts (34, 35). Additional studies of this mutation in familial prostate cancer are warranted to clarify its role.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: J.P. Breyer, J.R. Smith
Development of methodology: J.P. Breyer, J.R. Smith
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.P. Breyer, T.G. Avritt, K.M. McReynolds, J.R. Smith
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.P. Breyer, T.G. Avritt, W.D. Dupont, J.R. Smith
Writing, review, and/or revision of the manuscript: J.P. Breyer, W.D. Dupont, J.R. Smith
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.M. McReynolds, J.R. Smith
Study supervision: J.R. Smith
Acknowledgments
The authors thank the study participants and Drs. Sam Chang, Peter Clark, Michael Cookson, Rodney Davis, S. Duke Herrell, Richard Hock, William Maynard, Douglas Milam, Jason Pereira, and Joseph Smith.
Grant Support
This study was supported by a MERIT grant from the U.S. Department of Veteran's Affairs and by an award from the V Foundation.
