Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Jun 1.
Published in final edited form as: Eur Urol. 2014 Jan 4;65(6):1069–1075. doi: 10.1016/j.eururo.2013.12.058

Prostate Cancer (PCa) Risk Variants and Risk of Fatal PCa in the National Cancer Institute Breast and Prostate Cancer Cohort Consortium

Irene M Shui a,*, Sara Lindström a, Adam S Kibel b, Sonja I Berndt c, Daniele Campa d, Travis Gerke a, Kathryn L Penney a,e, Demetrius Albanes c, Christine Berg f, H Bas Bueno-de-Mesquita g,h,bb, Stephen Chanock c,i, E David Crawford j, W Ryan Diver k, Susan M Gapstur k, J Michael Gaziano a,l,m, Graham G Giles n,o, Brian Henderson p, Robert Hoover c, Mattias Johansson q,r, Loic Le Marchand s, Jing Ma a,e, Carmen Navarro t,u, Kim Overvad v, Fredrick R Schumacher p, Gianluca Severi n,w, Afshan Siddiq x, Meir Stampfer a,e, Victoria L Stevens k, Ruth C Travis y, Dimitrios Trichopoulos a,z,aa, Paolo Vineis w,bb, Lorelei A Mucci a, Meredith Yeager c,i, Edward Giovannucci a, Peter Kraft a
PMCID: PMC4006298  NIHMSID: NIHMS554719  PMID: 24411283

Abstract

Background

Screening and diagnosis of prostate cancer (PCa) is hampered by an inability to predict who has the potential to develop fatal disease and who has indolent cancer. Studies have identified multiple genetic risk loci for PCa incidence, but it is unknown whether they could be used as biomarkers for PCa-specific mortality (PCSM).

Objective

To examine the association of 47 established PCa risk single-nucleotide polymorphisms (SNPs) with PCSM.

Design, setting, and participants

We included 10 487 men who had PCa and 11 024 controls, with a median follow-up of 8.3 yr, during which 1053 PCa deaths occurred.

Outcome measurements and statistical analysis

The main outcome was PCSM. The risk allele was defined as the allele associated with an increased risk for PCa in the literature. We used Cox proportional hazards regression to calculate the hazard ratios of each SNP with time to progression to PCSM after diagnosis. We also used logistic regression to calculate odds ratios for each risk SNP, comparing fatal PCa cases to controls.

Results and limitations

Among the cases, we found that 8 of the 47 SNPs were significantly associated (p < 0.05) with time to PCSM. The risk allele of rs11672691 (intergenic) was associated with an increased risk for PCSM, while 7 SNPs had risk alleles inversely associated (rs13385191 [C2orf43], rs17021918 [PDLIM5], rs10486567 [JAZF1], rs6465657 [LMTK2], rs7127900 (intergenic), rs2735839 [KLK3], rs10993994 [MSMB], rs13385191 [C2orf43]). In the case-control analysis, 22 SNPs were associated (p < 0.05) with the risk of fatal PCa, but most did not differentiate between fatal and nonfatal PCa. Rs11672691 and rs10993994 were associated with both fatal and nonfatal PCa, while rs6465657, rs7127900, rs2735839, and rs13385191 were associated with nonfatal PCa only.

Conclusions

Eight established risk loci were associated with progression to PCSM after diagnosis. Twenty-two SNPs were associated with fatal PCa incidence, but most did not differentiate between fatal and nonfatal PCa. The relatively small magnitudes of the associations do not translate well into risk prediction, but these findings merit further follow-up, because they may yield important clues about the complex biology of fatal PCa. Patient summary: In this report, we assessed whether established PCa risk variants could predict PCSM. We found eight risk variants associated with PCSM: One predicted an increased risk of PCSM, while seven were associated with decreased risk. Larger studies that focus on fatal PCa are needed to identify more markers that could aid prediction.

Keywords: Prostate cancer, Risk single nucleotide polymorphisms, Prostate cancer mortality, Genetic epidemiology

1. Introduction

Prostate cancer (PCa) is one of the most heritable cancers, yet even though multiple studies have identified genetic markers for PCa incidence [1], few have investigated the association of genetic susceptibility with death from PCa. These data are urgently needed, because a large proportion of men diagnosed with PCa have an indolent form of disease and will die of causes other than PCa [2]. Determining biomarkers, such as single-nucleotide polymorphisms (SNPs), that can predict relevant clinical outcomes in men who have PCa—the most important being PCa-specific mortality (PCSM)—would provide potentially actionable information.

Only four studies have been conducted to date that specifically evaluated the association between established PCa risk loci and PCSM. Penney et al [3] investigated eight risk loci in the 8q24 and 17q regions among 6460 US PCa cases (693 PCa deaths) and found no significant associations. A Swedish study conducted by Wiklund et al [4] of 2875 PCa cases (440 PCa deaths), evaluated 16 risk loci, and found no significant associations. Gallagher et al [5] evaluated 29 risk loci in a study of 798 PCa cases (91 PCa deaths) of men with Ashkenazi Jewish ancestry and found that the risk allele of rs2735839 (KLK3) was inversely associated with PCSM. Pomerantz et al [6] assessed 35 risk loci in 3945 US PCa cases (580 PCa deaths). They observed an inverse association between PCSM and the risk allele of rs2735839 and an increased risk of PCSM with the risk allele of rs7676973, which was associated with an increased risk for PCSM. In the current study, we extend this work and assess whether a set of 47 risk variants were associated with the risk of fatal PCa and progression to PCSM in a large PCa cohort consortium.

2. Methods

2.1. Study population

The Breast and Prostate Cancer Cohort Consortium, a collaboration of eight cohort studies from Europe, the United States, and Australia, has been described in detail previously [7]. In each cohort, men who had incident PCa were identified through population-based cancer registries or self-reports confirmed by medical records, including pathology reports. Data on disease stage and grade at the time of diagnosis were collected from each cohort. Cases were followed for overall mortality and PCSM using a combination of death certificates, medical record review, and population registries. We restricted the current study to men who self-reported as being of European descent.

2.2. Single-nucleotide polymorphism selection and genotyping

We identified 55 confirmed PCa risk loci from the literature for which we had genotyping information. After excluding 8 SNPs in linkage disequilibrium (pairwise r2 > 0.2), 47 SNPs remained (Supplemental Table 1). Each risk SNP had been previously genotyped using the TaqMan assay (Applied Biosystems, Carlsbad, CA, USA) [8]. Blinded, duplicated samples indicated high-quality genotyping (>98.5% concordance). One autosomal SNP (rs1983891) did not meet Hardy-Weinberg equilibrium in the controls (p < 0.001).

2.3. Imputation

We imputed missing genotypes by sampling from the observed frequency distribution in men who had nonmissing genotype data in the same age category, stratified by case-control status (single conditional draw imputation) [9]. Genotypes from the Finnish site (ATBC) were imputed separately. Subjects from one site (MCCS) were excluded from the imputation because genotyping data for several of the risk SNPs were unavailable, and we also excluded subjects with >20% missing genotypes across the 47 SNPs. On average, <10% of the SNPs were imputed in this subset, and risk allele frequencies before and after imputations were similar; they are displayed in Supplemental Table 2.

2.4. Statistical methods

Fatal PCa was defined as men who had died of PCa. The risk allele for each SNP was defined as the allele associated with an increased risk of PCa in the literature. Among cases, we used Cox proportional hazards regression to calculate the per-risk allele hazard ratio (HR) and 95% confidence interval (CI) of each SNP with time to progression to PCSM after diagnosis. In this analysis, person-time to event began at the time of PCa diagnosis and ended at the date of PCSM; otherwise, men were censored if they died from other causes or the end of follow-up. The primary model adjusted for age at diagnosis and cohort, and a secondary model further adjusted for Gleason grade (2–7, 8–10) and clinical stage (A/B, C/D). The proportional hazards assumption was assessed using martingale residuals [10].

We also compared the distribution of the risk alleles in fatal PCa with nonfatal PCa using logistic regression. To determine whether the risk SNPs were associated with the incidence of fatal PCa, we used logistic regression to calculate the per-risk allele odds ratio (OR) and 95% CI for each risk SNP, comparing fatal PCa cases to controls. We then used polytomous logistic regression to assess whether the ORs for fatal PCa compared with controls differed from the ORs for nonfatal PCa compared with controls. For these analyses, we implemented a complete-case analysis to use all available data.

The cumulative association of the 47 PCa risk SNPs with progression to PCSM was assessed in two ways. We created an additive genetic score by summing the number of risk alleles (0, 1, 2) across the 47 SNPs for each individual. The median number of risk alleles carried was 42 (range: 27–59) in the cases. We calculated the HR for progression to PCSM by quintile of risk score, with the lowest quintile as the reference. To assess the joint multimarker association across the 47 SNPs, we used a kernel machine model with a linear kernel function [11,12]. In contrast to the risk allele score method, the kernel machine method does not require a priori assignment of directionality (eg, that SNPs associated with an increased risk of PCa incidence will also be associated with an increased risk of PCSM). Because these analyses could not accommodate missing data, we used the imputed SNP data. We report nominal two-sided p values without adjusting for multiple testing. All tests for significance were two-sided, and analyses were conducted in SAS v.9.3 software (SAS Institute, Cary, NC, USA) and R.

3. Results

Characteristics of the study population are described in Table 1. A total of 10 487 cases of PCa and 11 024 controls were included in the analysis. Among cases, the mean age at diagnosis was 68.6 yr of age. The majority of cases were localized stage (A/B) or low grade (2–7). Cases were followed for a median of 8.3 yr after cancer diagnosis; during this time, 1053 men died from PCa.

Table 1.

Study population characteristics

ATBC CPS-II EPIC HPFS MCCS MEC PHS PLCO All cohorts combined
Cases, no. 1036 2284 1573 1287 1031 665 1381 1230 10 487
% 10 22 15 12 10 6 13 12
Age at diagnosis, yr
  Mean (SD) 69.6 (5.7) 70.1 (5.7) 65.4 (6.3) 69.6 (7.5) 67.6 (6.9) 69.0 (7.5) 70.3 (7.6) 66.9 (5.5) 68.6 (6.8)
Grade, %
  2–7 64 72 46 76 83 65 83 90 72
  8–10 19 14 5 9 14 31 14 10 13
  Missing 17 14 49 15 3 4 3 0 15
Stage, %
  A or B 55 79 52 74 88 85 42 80 68
  C or D 25 18 16 11 9 13 17 20 16
  Missing 20 3 32 15 3 2 41 0 15
Year of diagnosis, %
  ≤1992 9 1 0 0 5 0 37 0 6
  1993–1997 37 33 11 34 24 30 27 35 29
  1998–2002 50 51 59 55 30 46 31 60 49
  2003–2007 4 15 30 11 40 24 6 6 16
Deaths
  Total deaths, no. 749 479 393 351 228 154 460 225 3039
  Total deaths, % 72 21 25 27 22 23 33 18 29
  PCa deaths, no. 289 126 213 82 75 28 176 64 1053
  PCa deaths, % 28 6 14 6 7 4 13 5 10
Controls, no. 952 2303 1845 1349 1397 743 1382 1053 11 024
% 9 21 17 12 13 7 13 10
Age at selection, yr
  Mean (SD) 69.0 (5.7) 70.2 (5.7) 65.4 (6.3) 67.6 (7.6) 54.4 (8.8) 70.4 (7.3) 70.2 (7.6) 69.0 (5.7) 66.8 (8.5)

SD = standard deviation; PCa = prostate cancer.

*

Percentages may not add up to 100% due to rounding.

3.1. Prostate cancer case–only analyses

Eight of the 47 risk loci were significantly (p < 0.05) associated with time to PCSM following diagnosis (Table 2; Supplemental Table 3). The risk allele for one SNP (rs11672691; Chr 19 intergenic) was associated with an increased rate of progression to PCSM (HR: 1.18; 95% CI, 1.05–1.34; p = 0.007). The other seven significantly associated SNPs (rs13385191, rs17021918, rs10486567, rs6465657, rs7127900, rs2735839, and rs10993994) had risk alleles that were inversely associated with PCSM (HRs ranging from 0.85 to 0.90). Adjusting for Gleason grade and clinical stage at diagnosis did not result in large changes in magnitude for most of the effect estimates, except for rs13385191, which was attenuated towards the null (Supplemental Table 3). Men categorized in the highest quintile of the additive risk score (eg, carrying the greatest number of risk alleles) showed a reduced risk of progression compared with those in the lowest quintile (HR: 0.77; 95% CI, 0.61–0.98), but the overall test for trend (p = 0.13) was not significant (Table 3). The global test for a combined effect across all risk loci showed a suggestive association with progression to PCSM (p = 0.05). Supplemental Table 4 shows the per-risk allele ORs and 95% CIs for each of the 47 SNPs in fatal compared with nonfatal cancers; the results are similar to those of the survival analysis.

Table 2.

Per-allele hazard ratios for risk single-nucleotide polymorphisms and time to prostate cancer–specific mortality*

SNP Gene Location Chr Risk
allele
Reference
allele
HR (95% CI) p value
rs13385191 C2orf43 Intronic 2 G A 0.88 (0.78–1.00) 0.05
rs17021918 PDLIM5 Intronic 4 C T 0.89 (0.81–0.97) 0.01
rs10486567 JAZF1 Intronic 7 G A 0.85 (0.76–0.94) 0.001
rs6465657 LMTK2 Intronic 7 C T 0.90 (0.82–0.98) 0.02
rs10993994 MSMB Upstream 10 T C 0.90 (0.83–0.98) 0.02
rs7127900 Intergenic 11 A G 0.86 (0.77–0.97) 0.01
rs11672691 N/A Intergenic 19 G A 1.18 (1.05–1.34) 0.007
rs2735839 KLK3 Downstream 19 G A 0.82 (0.73–0.93) 0.002

SNP = single-nucleotide polymorphism; Chr = chromosome; HR = hazard ratio; CI = confidence interval; PCa = prostate cancer.

*

10 487 men with PCa (1053 PCa-specific deaths); model adjusted for age and cohort.

Only findings with a p value <0.05 are presented in this table; for complete results, see Supplemental Table 3.

Table 3.

Association of single-nucleotide polymorphism score* (quintiles) with time to prostate cancer–specific mortality

Model 1^ Model 2^
HR 95% CI p trend HR 95% CI p trend
Quintile 1 1.00 Reference 1.00 Reference
Quintile 2 0.97 (0.76–1.23) 0.91 (0.72–1.16)
Quintile 3 0.93 (0.74–1.18) 0.13 0.95 (0.75–1.20) 0.3
Quintile 4 1.03 (0.83–1.27) 1.00 (0.81–1.24)
Quintile 5 0.77 (0.61–0.98) 0.82 (0.65–1.04)

HR = hazard ratio; CI = confidence interval.

*

The score equals the sum of the number of risk alleles over 47 risk SNPs.

^

Model 1 was adjusted for age and cohort; model 2 was additionally adjusted for Gleason grade and clinical stage.

3.2. Comparison of prostate cancer cases and controls

Supplemental Table 5 presents the ORs and 95% CIs for the association of each risk loci with fatal PCa compared with controls and nonfatal PCa compared with controls as well as the p value for heterogeneity (p contrast) for each of the 47 SNPs. We found that 22 of the risk SNPs were significantly associated with the risk of fatal PCa (p < 0.05) and that the magnitude and direction of the majority of the ORs for fatal PCa compared with controls were similar to those of nonfatal PCa compared with controls. Only one SNP (rs11672691) had a risk allele associated with a significantly larger risk of fatal PCa compared with controls (OR: 1.26; 95% CI, 1.11–1.43; p = 0.0005) than nonfatal PCa compared with controls (OR: 1.10; 95% CI, 1.05–1.16; p = 0.0003), with a p value for the contrast test <0.05. This was the same SNP whose risk allele was associated with an increased risk for progression to PCSM. When we compared the ORs for the other seven SNPs whose risk alleles were associated with a decreased risk of progression, we found that six of them (rs13385191, rs17021918, rs10486567, rs6465657, rs7127900, and rs2735839) were associated with nonfatal PCa only compared with controls (Table 4). Rs10993994 was significantly associated with fatal PCa compared with controls (OR: 1.11; 95% CI, 1.01–1.22; p = 0.03), but the association with nonfatal PCa compared with controls was stronger (OR: 1.24; 95% CI, 1.19–1.29; p < 0.0001; p contrast = 0.02).

Table 4.

Per-risk allele odds ratio for each risk single-nucleotide polymorphism and fatal prostate cancer (PCa) versus controls and nonfatal PCa versus controls*

Fatal Nonfatal
SNP Gene Location Chr Position Risk allele Reference
allele
OR (95% CI) p value OR (95% CI) p value p contrast
rs17021918 PDLIM5 Intronic 4 95781900 C T 0.97 (0.88–1.07) 0.55 1.09 (1.05–1.14) <0.0001 0.02
rs10486567 JAZF1 Intronic 7 27943088 G A 1.04 (0.94–1.16) 0.46 1.22 (1.16–1.28) <0.0001 0.004
rs6465657 LMTK2 Intronic 7 97654263 C T 0.98 (0.90–1.08) 0.73 1.11 (1.07–1.16) <0.0001 0.009
rs10993994 MSMB Upstream 10 51219502 T C 1.11 (1.01–1.22) 0.03 1.24 (1.19–1.29) <0.0001 0.02
rs7127900 Intergenic 11 2190150 A G 1.01 (0.89–1.13) 0.92 1.16 (1.10–1.22) <0.0001 0.02
rs11672691 N/A Intergenic 19 46677771 G A 1.26 (1.11–1.43) 0.0005 1.10 (1.05–1.16) 0.0003 0.05
rs2735839 KLK3 Downstream 19 56056435 G A 0.96 (0.84–1.09) 0.50 1.19 (1.12–1.26) <0.0001 0.001

SNP = single-nucleotide polymorphism; Chr = chromosome; OR = odds ratio; CI = confidence interval.

*

1053 fatal cases, 9434 nonfatal cases, 11 024 controls; model adjusted for age and study cohort. Only results with a p contrast <0.05 are shown; see Supplemental Table 5 for complete results.

4. Discussion

Few prior genetic risk studies have examined the most relevant end point of fatal PCa. With >10 000 men and 1053 PCa deaths, our study is the largest and most comprehensive to date to assess whether 47 established PCa risk variants are associated with PCSM. In our survival analysis, we found that eight SNPs predicted time to PCSM following diagnosis. We observed one SNP (rs11672691) whose risk allele was associated with both worse progression to PCSM following diagnosis and an increased risk of fatal PCa compared with controls. This SNP was also recently confirmed to be associated with aggressive PCa in a meta-analysis of several large genome-wide association studies (GWASs) [13]. Rs11672691 lies in the intergenic region on chromosome 19 between the genes ATP5SL and CEACAM21 and within LOC100505495, a hypothetical locus for a noncoding RNA [13]. The ATP5SL gene has been associated with height [14], but its function is unknown. CECAM21 belongs to the immunoglobulin superfamily of genes; genes in this family may have a role in cell adhesion and metastasis [15]. Additional examination of this region may yield further insight into the mechanisms behind PCa progression.

Interestingly, in our survival analysis, the risk alleles of seven SNPs were associated with a decreased risk of progression to PCSM following diagnosis. When we further investigated the association of these SNPs in cases and controls, we found that for the majority of these SNPs, this relationship was driven by the SNPs being associated with nonfatal PCa only. A potential explanation is that these SNPs could be markers of a factor that leads to earlier diagnosis or diagnosis of indolent disease. For example, the risk allele (G) of rs2735839 was inversely associated with time to PCSM, not associated with fatal PCa incidence compared with controls, and strongly associated with nonfatal PCa. The inverse association with time to PCSM has been observed in some [5,6] but not all other studies [4,16]. The G allele of rs2735839 has been associated with a decreased risk of high–Gleason grade cancers [17,18]. Rs2735839 is on chromosome 19 downstream from the KLK3 gene. KLK3 encodes prostate-specific antigen (PSA), and rs2735839-G has been associated with increased PSA levels in some studies [1921]. One hypothesis is that the protective effect of the risk allele (G) is mediated by a lead-time or ascertainment bias because of PSA screening. In a screened population, men carrying the risk allele who have higher PSA levels could have an increased biopsy rate, leading to more and earlier cancer diagnoses, while those without the risk allele would be less likely to be diagnosed with indolent cancer because of their lower PSA levels not reaching the threshold for biopsy. However, other studies have not seen an association with rs2735839 and PSA [6] or have found an association with the SNP and PCa risk even in cohorts whose cases were primarily not identified through PSA screening [17,20], consistent with a mechanism that is independent of PSA screening ascertainment bias.

Rs10993994 is located upstream from the gene MSMB on chromosome 10, which encodes a protein that is secreted by epithelial cells in the prostate. The risk allele of rs10993994 has been associated with higher PSA levels [4,19]. Thus, it is possible that PSA detection bias could also mediate some of the association with PCa. Independent of its influence on PSA, rs10993994 has been associated with RNA expression of the MSMB and NCOA4 genes, which may mediate prostate carcinogenesis through transforming prostate cells to become anchorage independent [22]. We found that rs10993994 was associated with fatal PCa compared with controls, but the association was stronger for nonfatal PCa. Also, the risk allele was more common in nonfatal PCa compared with fatal PCa and was associated with a decreased rate of progression to PCSM. Ahn et al [23] also observed that the risk allele was associated with an increased risk for metastatic PCa compared with controls but was not associated with time to recurrence in PCa cases following diagnosis. A study comparing aggressive and nonaggressive cases of PCa found the risk allele to be more frequent in men who had less aggressive disease [18]. Three case-only studies did not find an association with rs10993994 and time to PCSM [46].

Rs10486567, which lies in an intron of the JAZF1 gene on chromosome 7, had the strongest association with time to PCSM (HR: 0.85; 95% CI, 0.76–0.94; p = 0.001). The risk allele of this SNP was also associated with a significantly decreased risk of biochemical recurrence and clinical metastases as well as a nonsignificantly decreased risk of PCSM in one study [5]. However, two other case-only studies found no association with PCSM [4,6]. The biological function of this SNP is unclear and merits further investigation.

The other nominally significant SNPs that were associated with time to PCSM following diagnosis were rs13385191, rs17021918, rs6465657, and rs7127900. Both rs17021918 (intronic in PDLIM5; Chr 4) and rs7127900 (intergenic; Chr 11) were assessed in the Pomerantz et al study [6], and there was no significant association for either SNP with PCSM; these SNPs were not assessed in other studies [4,5]. The Swedish case-only study [4] also observed a nominally significant inverse association with the risk allele of rs6465657 (intronic in LMTK2; Chr 7) and PCSM, but another United States–based study did not find an association [6]. To our knowledge, Rs13385191 (intronic in C2orf43; Chr 2) has not been investigated for an association with PCSM in other studies.

The development of prognostic tools to determine progression to PCSM is a much-needed and active area of research. A few promising models exist [24,25] that incorporate factors such as stage, Gleason grade, and PSA, but they are in need of external validation. Our study sample was limited because we did not have consistent access to more granular information on stage, grade, and PSA at diagnosis as well as other clinicopathologic information (eg, tumor volume) and treatment information that would be useful for testing prediction. Even so, we did not observe a change in C-statistic when we compared a Cox proportional hazards model predicting time to PCSM that included age at diagnosis, Gleason grade, and stage (c = 0.777) against a model with those variables and the additional eight SNPs that were predictive of time to PCSM (c = 0.781), indicating that the modest associations of this subset of risk SNPs do not translate well into clinical prediction.

Finally, our case-control results showed that most of the 47 established risk alleles were associated with both fatal and nonfatal PCa in a similar manner; 22 of the 47 SNPs were significantly associated with fatal PCa, but the majority of the SNPs did not differentiate between fatal and nonfatal disease. It is likely that many of these SNPs are necessary for the initiation of cancer, and other markers or exposures determine progression. The one exception was rs11672691, which was more strongly associated with fatal disease and also predicted PCSM following diagnosis.

An important point is that this study focused on known risk variants of PCa incidence and was also limited to the 47 SNPs for which we had genotyping data. Thus, it was not designed to identify novel variants. GWASs specifically designed to assess PCSM are needed to identify further genetic markers that could improve predictive ability, although to date, few studies have been large enough to do so [26]. New risk loci continue to be confirmed that were not included in our study. For example, the most recently published study has identified 23 novel PCa risk loci, including 16 that were associated with aggressive disease; however, these SNPs did not differentiate between aggressive and nonaggressive PCa when compared with controls [1].

Most genetic studies have relied on Gleason grade as a surrogate of PCa aggressiveness, but not all high–Gleason grade cancers will progress to PCSM. In addition, although Gleason grade is strongly associated with progression, the positive predictive value of Gleason score is relatively low (<30% for Gleason ≥7) [27,28], making it a suboptimal proxy. A major strength of our study was that we were able to follow men over time and accrue >1000 deaths from PCa. Still, with longer follow-up, it is possible that some of the nonfatal cases will progress to PCSM; thus, there may be some misclassification of our end point that would reduce our power to detect effects, especially when comparing fatal and nonfatal cases. Likewise, misclassification of PCSM resulting from the use of death registries as primary sources for this information may exist. Several of the contributing sites had dedicated end points committees to determine PCSM based on the death registries combined with information from medical records as well as physician and kin reporting to reduce this misclassification. Despite our relatively large sample size, it is possible that our findings could be the result of chance, but our finding that 8 of the 47 SNPs were associated with time to PCSM was more than the 2.4 loci that would be expected by chance (p = 0.002). Moreover, we observed a borderline significant (p = 0.05) global association across all of the risk loci with time to PCSM, supporting idea that our findings were not the result of chance.

5. Conclusions

Although dozens of common germline genetic risk variants are clearly associated with overall PCa risk, a key question with high clinical utility is whether these markers can improve the prognostication of PCSM following diagnosis. We identified eight SNPs that were associated with time to PCSM, including rs11672691, whose risk allele was predictive of increased progression to PCSM and increased risk of fatal PCa. The associations observed were of fairly small magnitude and thus do not translate well into improved risk prediction, but these findings merit further follow-up to investigate the biological mechanisms behind the associations. Future GWASs focused on fatal PCa are needed to identify novel markers that can differentiate disease aggressiveness and therefore be integrated into clinical practice and help clarify the biology behind PCa progression.

Supplementary Material

01

Take-home message.

Although several germline genetic risk variants have been established for prostate cancer (PCa) incidence, a key question is whether they are also related to survival. We assessed 47 PCa risk loci and found a subset that was related to PCa mortality.

Acknowledgments

Financial disclosures: Irene M. Shui certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/ affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Drs. Shui and Gerke were supported by a National Research Service Award (T32 CA09001) from the US National Institutes of Health (NIH) National Cancer Institute (NCI). Dr. Shui also received a US Army Department of Defense Prostate Cancer Postdoctoral Fellowship. Drs. Penney and Mucci were supported by the Prostate Cancer Foundation.

Funding/Support and role of the sponsor: This work was supported by the NIH NCI (cooperative agreement U19 CA148537-01). The maintenance of the Cancer Prevention Study II is supported by the American Cancer Society, and genotyping of the CPS-II samples was supported by a grant from the NCI (5U01CA098710). The Danish study Diet, Cancer and Health was funded by the Danish Cancer Society. EPIC-Greece was supported through the Hellenic Health Foundation. EPIC-Spain was supported by Health Research Fund; the regional governments of Andalucía, Asturias, Basque Country, Murcia (No. 6236), and Navarra; and ISCIII RETIC (RD06/0020; Spain). PLCO was supported by the intramural program of the Division of Cancer Epidemiology and Genetics, NCI. The Melbourne Collaborative Cohort Study recruitment was funded and its follow-up supported by Cancer Council Victoria.

Acknowledgment statement: The authors acknowledge Hongyan Huang for assistance in preparing the data.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Author contributions: Irene M. Shui had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Shui, Lindström, Kraft.

Acquisition of data: Shui, Lindström, Giles, Henderson, Le Marchand, Schumacher, Severi, Berndt, Campa, Albanes, Berg, Bueno-de-Mesquita, Chanock, Crawford, Diver, Gapstur, Gaziano, Hoover, Johansson, Ma, Navarro, Overvad, Siddiq, Stampfer, Stevens, Travis, Trichopoulos, Vineis, Yeager, Kraft.

Analysis and interpretation of data: Shui, Lindström, Berndt, Campa, Gerke, Penney, Kibel, Mucci, Giovannucci, Kraft.

Drafting of the manuscript: Shui, Lindström, Berndt, Campa, Gerke, Kibel, Penney, Giovannucci, Kraft.

Critical revision of the manuscript for important intellectual content: Shui, Lindström, Kibel, Berndt, Campa, Albanes, Giles, Henderson, Le Marchand, Schumacher, Severi, Berg, Buenode-Mesquita, Chanock, Crawford, Diver, Gapstur, Gaziano, Hoover, Johansson, Ma, Navarro, Overvad, Siddiq, Stampfer, Stevens, Travis, Trichopoulos, Vineis, Yeager, Kraft, Mucci, Giovannucci, Gerke, Penney.

Statistical analysis: Shui, Lindström, Gerke, Kraft.

Obtaining funding: Kraft.

Administrative, technical, or material support: None.

Supervision: Kraft, Lindström, Giovannucci.

Other (specify): None.

References

  • 1.Eeles RA, Olama AA, Benlloch S, et al. COGS–Cancer Research UK GWAS–ELLIPSE (part of GAME-ON) Initiative; Australian Prostate Cancer Bioresource; UK Genetic Prostate Cancer Study Collaborators/British Association of Urological Surgeons’ Section of Oncology; UK ProtecT (Prostate testing for cancer and Treatment) Study Collaborators; PRACTICAL (Prostate Cancer Association Group to Investigate Cancer-Associated Alterations in the Genome) Consortium. Identification of 23 new prostate cancer susceptibility loci using the iCOGS custom genotyping array. Nat Genet. 2013;45:385–391. doi: 10.1038/ng.2560. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Epstein MM, Edgren G, Rider JR, Mucci LA, Adami HO. Temporal trends in cause of death among Swedish and US men with prostate cancer. J Natl Cancer Inst. 2012;104:1335–1342. doi: 10.1093/jnci/djs299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Penney KL, Salinas CA, Pomerantz M, et al. Evaluation of 8q24 and 17q risk loci and prostate cancer mortality. Clin Cancer Res. 2009;15:3223–3230. doi: 10.1158/1078-0432.CCR-08-2733. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Wiklund FE, Adami HO, Zheng SL, et al. Established prostate cancer susceptibility variants are not associated with disease outcome. Cancer Epidemiol Biomarkers Prev. 2009;18:1659–1662. doi: 10.1158/1055-9965.EPI-08-1148. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Gallagher DJ, Vijai J, Cronin AM, et al. Susceptibility loci associated with prostate cancer progression and mortality. Clin Cancer Res. 2010;16:2819–2832. doi: 10.1158/1078-0432.CCR-10-0028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Pomerantz MM, Werner L, Xie W, et al. Association of prostate cancer risk loci with disease aggressiveness and prostate cancer-specific mortality. Cancer Prev Res (Phila) 2011;4:719–728. doi: 10.1158/1940-6207.CAPR-10-0292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Hunter DJ, Riboli E, Haiman CA, et al. National Cancer Institute Breast and Prostate Cancer Cohort Consortium. A candidate gene approach to searching for low-penetrance breast and prostate cancer genes. Nat Rev Cancer. 2005;5:977–985. doi: 10.1038/nrc1754. [DOI] [PubMed] [Google Scholar]
  • 8.Yeager M, Orr N, Hayes RB, et al. Genome-wide association study of prostate cancer identifies a second risk locus at 8q24. Nat Genet. 2007;39:645–649. doi: 10.1038/ng2022. [DOI] [PubMed] [Google Scholar]
  • 9.Lindström S, Schumacher FR, Cox D, et al. Common genetic variants in prostate cancer risk prediction—results from the NCI Breast and Prostate Cancer Cohort Consortium (BPC3) Cancer Epidemiol Biomarkers Prev. 2012;21:437–444. doi: 10.1158/1055-9965.EPI-11-1038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Lin DY, Wei LJ, Ying Z. Checking the Cox model with cumulative sums of martingale-based residuals. Biometrika. 1993;80:557–572. [Google Scholar]
  • 11.Lin X, Cai T, Wu MC, Zhou Q, Liu G, Christiani DC. Kernel machine SNP-set analysis for censored survival outcomes in genome-wide association studies. Genet Epidemiol. 2011;35:620–631. doi: 10.1002/gepi.20610. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Wu MC, Kraft P, Epstein MP, et al. Powerful SNP-set analysis for case-control genome-wide association studies. Am J Hum Genet. 2010;86:929–942. doi: 10.1016/j.ajhg.2010.05.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Amin Al Olama A, Kote-Jarai Z, Schumacher FR, et al. UK Genetic Prostate Cancer Study Collaborators/British Association of Urological Surgeons’ Section of Oncology; UK ProtecT Study Collaborators; Australian Prostate Cancer Bioresource; PRACTICAL Consortium. A meta-analysis of genome-wide association studies to identify prostate cancer susceptibility loci associated with aggressive and non-aggressive disease. Hum Mol Genet. 2013;22:408–415. doi: 10.1093/hmg/dds425. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Lango Allen H, Estrada K, Lettre G, et al. Hundreds of variants clustered in genomic loci and biological pathways affect human height. Nature. 2010;467:832–838. doi: 10.1038/nature09410. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Blumenthal RD, Leon E, Hansen HJ, Goldenberg DM. Expression patterns of CEACAM5 and CEACAM6 in primary and metastatic cancers. BMC Cancer. 2007;7:2. doi: 10.1186/1471-2407-7-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Penney KL, Schumacher FR, Kraft P, et al. Association of KLK3 (PSA) genetic variants with prostate cancer risk and PSA levels. Carcinogenesis. 2011;32:853–859. doi: 10.1093/carcin/bgr050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Lindstrom S, Schumacher F, Siddiq A, et al. Characterizing associations and SNP-environment interactions for GWAS-identified prostate cancer risk markers—results from BPC3. PLoS One. 2011;6:e17142. doi: 10.1371/journal.pone.0017142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Kader AK, Sun J, Isaacs SD, et al. Individual and cumulative effect of prostate cancer risk-associated variants on clinicopathologic variables in 5,895 prostate cancer patients. Prostate. 2009;69:1195–1205. doi: 10.1002/pros.20970. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Ahn J, Berndt SI, Wacholder S, et al. Variation in KLK genes, prostate-specific antigen and risk of prostate cancer. Nat Genet. 2008;40:1032–1034. doi: 10.1038/ng0908-1032. author reply 1035–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Eeles RA, Kote-Jarai Z, Giles GG, et al. Multiple newly identified loci associated with prostate cancer susceptibility. Nat Genet. 2008;40:316–321. doi: 10.1038/ng.90. [DOI] [PubMed] [Google Scholar]
  • 21.Gudmundsson J, Besenbacher S, Sulem P, et al. Genetic correction of PSA values using sequence variants associated with PSA levels. Sci Transl Med. 2010;2:62ra92. doi: 10.1126/scitranslmed.3001513. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Pomerantz MM, Shrestha Y, Flavin RJ, et al. Analysis of the 10q11 cancer risk locus implicates MSMB and NCOA4 in human prostate tumorigenesis. PLoS Genet. 2010;6:e1001204. doi: 10.1371/journal.pgen.1001204. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Ahn J, Kibel AS, Park JY, et al. Prostate cancer predisposition loci and risk of metastatic disease and prostate cancer recurrence. Clin Cancer Res. 2011;17:1075–1081. doi: 10.1158/1078-0432.CCR-10-0881. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.van Leeuwen PJ, van den Bergh RC, Wolters T, et al. Critical assessment of prebiopsy parameters for predicting prostate cancer metastasis and mortality. Can J Urol. 2011;18:6018–6024. [PubMed] [Google Scholar]
  • 25.Nguyen PL, Chen MH, Catalona WJ, Moul JW, Sun L, D’Amico AV. Predicting prostate cancer mortality among men with intermediate to high-risk disease and multiple unfavorable risk factors. Int J Radiat Oncol Biol Phys. 2009;73:659–664. doi: 10.1016/j.ijrobp.2008.05.009. [DOI] [PubMed] [Google Scholar]
  • 26.Penney KL, Pyne S, Schumacher FR, et al. Genome-wide association study of prostate cancer mortality. Cancer Epidemiol Biomarkers Prev. 2010;19:2869–2876. doi: 10.1158/1055-9965.EPI-10-0601. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Andren O, Fall K, Franzen L, Andersson SO, Johansson JE, Rubin MA. How well does the Gleason score predict prostate cancer death? A 20-year followup of a population based cohort in Sweden. J Urol. 2006;175:1337–1340. doi: 10.1016/S0022-5347(05)00734-2. [DOI] [PubMed] [Google Scholar]
  • 28.Shui IM, Mucci LA, Kraft P, et al. Vitamin D–related genetic variation, plasma vitamin D, and risk of lethal prostate cancer: a prospective nested case-control study. J Natl Cancer Inst. 2012;104:690–699. doi: 10.1093/jnci/djs189. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01

RESOURCES