A systematic review of the prevalence of DNA damage response gene mutations in prostate cancer.

Several ongoing international prostate cancer (PC) clinical trials are exploring therapies that target the DNA damage response (DDR) pathway. This systematic review summarizes the prevalence of DDR mutation carriers in the unselected (general) PC and familial PC populations. A total of 11 electronic databases, 10 conference proceedings, and grey literature sources were searched from their inception to December 2017. Studies reporting the prevalence of somatic and/or germline DDR mutations were summarized. Metastatic PC (mPC), castration‑resistant PC (CRPC) and metastatic CRPC (mCRPC) subgroups were included. A total of 11,648 records were retrieved, and 80 studies (103 records) across all PC populations were included; 59 records were of unselected PC and 13 records of familial PC. Most data were available for DDR panels (n=12 studies), ataxia telangiectasia mutated (ATM; n=13), breast cancer susceptibility gene (BRCA)1 (n=14) and BRCA2 (n=20). ATM, BRCA2 and partner and localizer of BRCA2 (PALB2) had the highest mutation rates (≥4%). Median prevalence rates for DDR germline mutations were 18.6% in PC (range, 17.2‑19%; three studies, n=1,712), 11.6% in mPC (range, 11.4‑11.8%; two studies, n=1,261) and 8.3% in mCRPC (range, 7.5‑9.1%; two studies, n=738). Median prevalence rates for DDR somatic mutations were 10.7% in PC (range, 4.9‑22%; three studies, n=680), 13.2% in mPC (range, 10‑16.4%; two studies, n=105) and not reported (NR) in mCRPC. The prevalence of DDR germline and/or somatic mutations was 27% in PC (one study, n=221), 22.67% in mCRPC (one study, n=150) and NR in mPC. In familial PC, median mutation prevalence was 12.1% (range, 7.3‑16.9%) for germline DDR (two studies, n=315) and 3.7% (range, 1.3‑7.9%) for BRCA2 (six studies, n=945). In total, 88% of studies were at a high risk of bias. The prevalence of DDR gene mutations in PC varied widely within somatic subgroups depending on study size, genetic screening techniques, DDR mutation definition and PC diagnosis; somatic and/or germline DDR mutation prevalence was in the range of 23‑27% in PC. These findings support DDR mutation testing for all patients with PC (including those with mCRPC). With the advent of the latest clinical practice PC guidelines highlighting the importance of DDR mutation screening, and ongoing mCRPC clinical trials evaluating DDR mutation‑targeted drugs, future larger epidemiological studies are warranted to further quantify the international burden of DDR mutations in PC.

Introduction

Prostate cancer (PC) is a major global health burden, with a worldwide incidence of 1.1 million in 2012 (1). While typical active treatments may involve surgery, chemotherapy, brachytherapy and/or androgen-deprivation therapy, these often fail to be curative, with many patients developing metastatic disease or therapeutic resistance. Aggressive prostate disease, such as metastatic castration-resistant PC (mCRPC), is usually lethal.

The DNA damage response (DDR) is an essential pathway that ensures the survival of both normal and malignant prostate cells and includes many important genes, such as breast cancer susceptibility gene (BRCA)1/2, ataxia telangiectasia mutated (ATM) and partner and localizer of BRCA2 (PALB2) (2). Efficient and specific repair of DNA damage maintains the genomic integrity of the cell and ensures its ability to persist and proliferate. Mutations in DDR genes contribute to destabilizing PC cells, often making them more susceptible to cell death (3). Poly (ADP)-ribose polymerase (PARP) inhibitors represent an emerging therapeutic approach to target the DDR pathway in malignant cells (3). In cancer cells that already harbor multiple other genetic mutations (e.g. BRCA1/2 mutations), PARP inhibition can render the cell unable to repair DNA damage, leading to cell death (3).

In PC, mutations in genes involved in the DDR pathway are relatively common, particularly in the advanced stages of the disease (4). Several ongoing international PC clinical trials are exploring PARP inhibitors that target the DDR pathway, and the contribution of DDR gene mutations to improving therapeutic outcomes in the context of PARP inhibitor therapy is becoming more evident (Swift et al, unpublished data).

Treatment guidelines and consensus statements from international clinical organizations have recently been published to reflect the importance of DDR mutation screening for the management of PC, although the precise genes or gene panels are not always in accordance. The National Comprehensive Cancer Network (NCCN) 2018 guidelines for PC state that a strong family history consists of: Brother, father or multiple family members diagnosed with prostate cancer by at least 60 years of age; and known germline DNA repair gene abnormalities, especially BRCA2 mutation. These guidelines advise clinicians as follows: 'Consider testing for mutations in these genes (germline and somatic): BRCA1, BRCA2, ATM, PALB2, FANCA; refer to genetic counseling if positive. At present, this information may be useful for genetic counseling, early use of platinum chemotherapy, or eligibility for clinical trials (e.g. PARP inhibitors)' (5). The Philadelphia Prostate Cancer Consensus Conference 2017 recommended that all patients with familial and metastatic PC (mPC) consider genetic testing [encompassing ATM, BRCA1, BRCA2, nibrin (NBN) and DNA mismatch repair genes] (6): 'There was strong consensus to factor BRCA2 mutations into [prostate cancer] screening discussions. BRCA2 achieved moderate consensus for factoring into early stage management discussion, with stronger consensus in high-risk/advanced and metastatic setting[s]' (6). The St Gallen Advanced Prostate Cancer Consensus Conference 2017 reported 'that BRCA1, BRCA2 and ATM mutations should be reported [for mCRPC] because that knowledge will likely influence management decisions' (7). It is important that national/regional healthcare providers and decision makers be kept informed of the burden of DDR mutations in PC.

The authors of the present study undertook a systematic review of data to identify and summarize the prevalence of DDR mutations in the unselected (general) population of patients with PC (including mPC, mCRPC and CRPC). Selected subgroup data for familial PC were also presented, since this population is currently considered a key focus for genetic testing guidelines. Studies that reported on other selected subgroups (including young-onset PC, lethal PC, ductal PC, patients receiving pre-specified treatment regimens, and Ashkenazi Jewish and African-American populations) were identified but were not the focus of the review. Methodological factors and limitations of the included studies that may have led to variation in the prevalence rates reported are noted and discussed.

Materials and methods

Inclusion criteria

This systematic review was carried out in accordance with the methodologies recommended by the Cochrane Collaboration (8) and the Centre for Reviews and Dissemination (9). The review adhered to a pre-defined protocol, which stipulated the methodology provided below. Studies that reported the prevalence of mutated DDR genes in men with PC, castration-resistant PC (CRPC), mPC or mCRPC were included. Detailed inclusion criteria (including DDR definition) are provided in Data S1, Appendix S1.

Search methods

In order to identify relevant studies, a range of electronic databases (n=11) were searched from their inception to December 2017, including Medline (Ovid), Embase (Ovid), CINAHL (EBSCO) and the Cochrane Database of Systematic Reviews (Wiley). Searches used a combination of text and database thesaurus terms. No restrictions on language or publication status were applied. Searches of conference abstracts and reference lists of articles were conducted. Full details of the search methods employed, the databases searched, and the Embase search strategy are provided in Data S1, Appendix S2. Titles and abstracts were independently screened by two reviewers. Full paper copies were independently examined in detail by two reviewers to determine whether or not they met the inclusion criteria. Data were extracted from the included studies using a specifically designed and piloted data extraction sheet developed using Microsoft Excel 2016 (Microsoft Corporation). Details of the study methods, population characteristics, risk of bias, and outcome data were extracted from each study by one reviewer and checked for accuracy against the original publication by a second reviewer. Criteria used to assess the risk of bias were taken from the Joanna Briggs Institute Critical Appraisal Checklist for studies reporting prevalence data (10). Any disagreements or discrepancies in study selection or data extraction were resolved by consensus or through consultation with a third reviewer.

Data synthesis

A narrative summary of all included studies was produced. The results prioritized the following countries: Australia, Canada, France, Germany, Israel, Italy, Japan, Korea, Russia, Spain, the UK and the USA. Multinational studies were summarized separately. Analysis of prevalence was primarily conducted for the unselected PC population and familial PC subgroups only; data for additional subgroups identified during screening (including young-onset PC, ductal PC, lethal PC, African-American with PC, Ashkenazi Jewish with PC or patients receiving pre-specified treatment regimens) were also extracted. Studies were excluded from the analysis of prevalence if they involved: i) Specific mutation(s) in a given gene rather than all mutations for a given gene (unless the study was familial); ii) <50 participants (unless the study was familial), to focus on well powered studies (11,12); or iii) did not clearly report whether a mutation was germline or somatic. Such studies are summarized in Data S1, Appendix S3, together with reported prevalence values. The median prevalence and range were reported for each gene of interest or combination of genes (DDR). If multiple definitions were available for a given gene mutation, then the broadest definition was included where possible. The median prevalence rates were compared between germline and somatic mutations or between prostate subgroups if the dataset was greater than 500 combined participants, to focus on data with a large sample size.

Results

Search methods and inclusion assessment

In total, 11,648 titles were retrieved from the database searches. A total of 14 articles were identified from hand and citation searching. Four articles were identified from conference abstracts. Fig. 1 summarizes the flow of studies through the search and screening process. A total of 265 articles were excluded after the full records were read (Fig. 1 and Data S1, Appendix S3). There were 123 records identified for inclusion in the review, though 20 records were not from countries of interest and were not considered further (Data S1, Appendix S3). For the prioritized countries, 103 records from 80 studies (some of which had multiple records or publications) were identified for inclusion in the present review. A total of 59 records focused on unselected patient populations, and 13 records focused on familial PC patients.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-analyses study flow diagram. DDR, DNA damage response; ESMO, European Society for Medical Oncology; NCCN, National Comprehensive Care Network; SUO, Society of Urologic Oncology.

Several records reporting on additional selected subgroups were identified, including 11 for Ashkenazi Jewish patients, 10 for patients receiving pre-specified treatment regimens, five for young-onset (≤55 or 65 years) PC, five for lethal PC, two for ductal PC and one for African-American patients with PC. The results for these subgroups are presented in Data S1, Appendices S3 and S4, together with reported prevalence values.

Unselected populations

A total of 47 studies (59 records) were identified for the unselected population. Out of the 47 studies, the majority (22 studies) were conducted exclusively in the USA (Table I). A total of five studies were conducted in the UK, three in Canada and two in Japan. Denmark, Germany, Israel and Spain each provided a single study. A total of nine studies recruited patients from multiple locations; the country of recruitment was not reported for two studies (both abstracts).

Table I

Study characteristics: Unselected populations.

PC group		Study ID	Other related publications	No. analyzed	Recruitment/enrollment dates	Place(s) of recruitment
A, Canada
PC		Akbari et al, 2014 (36)	NR	1904	June 1998-February 2010	Sunnybrook Health Sciences Center, Toronto, Canada; Princess Margaret Hospital, Toronto, Canada
		Nam et al, 2005 (57)	NR	996	June 1998-January 2003	Two prostate centers at the University of Toronto (University Health Network and Sunnybrook and Women's College Health Sciences Center)
mCRPC		Struss et al, 2017a,(23)	NR	319	NR	NR
B, Denmark
PC		Naslund-Koch et al, 2016 (58)	NR	39014	2003-2010	Copenhagen General Population Study
C, Germany
PC		Maier et al, 2014 (37)	Maier 2010 (59)	382; 92	1998-2007	Ulm, Germany
D, Israel
PC		Vazina et al, 2000 (60)	NR	174	1998	Rabin Medical Center, Sheba Medical Center or Wolfson Medical Center
E, Japan
PC		Tanaka et al, 2009 (61)	NR	177	1997-2003	Shimane Medical University, Izumo, Japan
		Uchida et al, 1999 (62)	NR	24	NR	University of Kitasato Hospital, Sagamihara, Kanagawa, Japan
F, Spain
mCRPC		Romero Laorden et al, 2017a (22) NCT03075735 (PROREPAIR-B)	Cendón Flórez et al, 2017 (63); NCT3075735 (64)	419	NR	38 centers across Spain
G, UK
PC		Castro et al, 2011a (31) UKGPCS	NR	2181	1990-2005	UK Genetic Prostate Cancer Study (UKGPCS)
PC		Manson-Bahr et al, 2015 (27)	NR	63	NR	NR
PC		Leongamornlert et al, 2012 (32) UKGPCS	NR	886	NR	UK Genetic Prostate Cancer Study (UKGPCS)
PC		Angele et al, 2004 (65)	NR	637	1993-2002	Royal Marsden NHS Trust
PC		Jefferies et al, 2017a (14)	NR	61	NR	Welsh Cancer Bank
H, USA
PC		Dawson et al, 2016a (66)	NR	437	NR	NR
PC		Nicolosi et al, 2017a (19)	NR	1158	2013-2016
PC	Xia et al, 2015 (67)		NR	20	NR	NR; hospital-based registry
PC	Abida et al, 2017 (17)		Abida et al, 2015 (68);	451	May 2015-unknown	Memorial Sloan Kettering Cancer Center, New York, USA
			Abida et al, 2016 (69);
			Abida et al,2016 (70);
			Cheng et al, 2017 (71)
PC	Feldman et al, 2014a (72)		NR	330	NR	Large referral laboratory
PC	Myers et al, 2016a (35)		NR	85	NR	NR
PC	Palapattu et al, 2015a (73)		NR	9	NR	NR
PC	Browning et al, 2006 (74)		NR	98	1997	Vanderbilt Hospital, USA
PC	Patel et al, 2016a (29)		Patel et al, 2017 (75)	327	NR	Dana Farber Cancer Institute
PC	Lara et al, 2017a (34)		NR	207; 936	NR	NR
PC	Evans et al, 2016 (28)		NR	1090	NR	Mayo Clinic, Cleveland Clinic, Thomas Jefferson University
PC	Wu et al, 2006 (76)		NR	84	1997-1998	Mayo Clinic, USA
PC	Williams et al, 1996 (47)		Gao et al, 1995 (48)	23	NR	University of Utah
Primary PC;	Grasso et al, 2012 (26)		NR	11; 50	NR	University of Michigan and from the Rapid Autopsy
mCRPC						Program
Primary PC; mPC	Dall'Era et al, 2017a (77)		Glass et al, 2017 (78)	936	NR	NR
Primary PC; mPC	Beltran et al, 2015 (79)		Beltran et al, 2015 (80)	69; 29	February 2013- September 2014	Weill Cornell Medical College–New York Presbyterian Hospital
Primary PC; mPC;	Beltran et al, 2013 (81)		NR	16; 4; 25	NR	Weill Cornell Medical College
CRPC
mCRPC	Gambhira et al, 2016a (82)		NR	13	NR	NR
mCRPC	Daniel et al, 2017a (83)		NR	1911	NR	NR
mPC	Pritchard et al, 2014 (16)		NR	60	NR	University of Washington Prostate Cancer Biorepository
mPC	Gourdin and Lilly 2016a (15)		NR	55	NR	NR
mPC	Robbins et al, 2011 (84)		NR	8	NR	Rapid Autopsy Program at the University of Michigan
I, Multi-national
PC	Fontugne et al, 2015a (38)		NR	51	NR	UK and USA
PC	Cancer Genome Atlas 2015 (18)		NR	333	NR	Multiple institutions (including Australia, Brazil, Israel, UK, USA)
PC	Lu et al, 2015 (24) Cancer Genome Atlas		NR	178	NR	NR, but presumed to be multiple institutions (including Australia, Brazil, Israel, UK, USA) (18)
PC	Timms et al, 2016c (13)		NR	84	1990-2011	Transatlantic Prostate Group Cohort and commercial databank
PC	Hebbring et al, 2006 (85)		NR	1819; 1218	NR	USA (Johns Hopkins, Tampere University Hospital, University of Michigan, Mayo Clinic); Germany (Universitätsklinikum Ulm)
PC; mCRPC	Decker et al, 2016 (30)		Baca et al, 2013 (86)	60; 150	NR	Multiple institutions (including Australia, Brazil, Israel, UK, and USA); unclear for some institutions
Primary PC; mPC	Pritchard et al, 2016 (21) Stand Up to Cancer-Prostate Cancer Foundation (SU2C-PCF) International Prostate Cancer Dream Team discovery series; MSK-IMPACT		NR	499; 692	1997-2015 (multiple case series)	Royal Marsden Hospital (London, UK), University of Washington Rapid Autopsy Program (District of Columbia, USA), Weill Cornell Medical College (New York, USA), University of Michigan Rapid Autopsy Program (Michigan, USA), Memorial Sloan Kettering Cancer Center (New York, USA); 144 international centers (SU2C-PCF)
mCRPC	Robinson et al, 2015 (4)		NR	150	NR	8 centers, including University of Michigan Medical School, University of Washington, Dana-Farber Cancer Institute, Memorial Sloan Kettering Cancer Center, Royal Marsden Institute of Cancer Research (London). Karmanos Cancer Institute, Beth Israel Deaconess Medical Center, Harvard Medical School, and Weill-Cornell
mPC	Nelson et al, 2016a (20)		NR	569	NR	NR but includes UK and USA
J, Country not reported
PC	Liu et al, 2016a (87)		NR	36	NR	NR
mCRPC	Sonpavde et al, 2017a (33)		NR	514	NR	NR

Abstract;

letter;

poster.

Note that some unselected populations could also present data for the selected populations. CRPC, castration resistant prostate cancer; mCRPC, metastatic castration resistant prostate cancer; mPC, metastatic prostate cancer; NR, not reported; PC, prostate cancer.

In total, 37 of the 47 primary studies reported results for patients with unselected PC, five of which focused on primary or localized PC (the authors' descriptions are reported where the definitions of disease terms were not made more explicit). Four studies focused on mPC and six studies focused on mCRPC.

The numbers of patients in the studies varied considerably, ranging from 8 to 39,014. A total of 10 studies included <50 patients in their cohort, 23 studies included between 50 and 500 patients, and 14 studies included >500 patients. The majority of studies failed to report recruitment or enrollment dates (32 studies). Of the remaining studies, 11 commenced recruitment between 1990 and 1998, and only four recruited patients after the year 2000.

Detailed patient baseline details are presented in Data S1, Appendix S5. Generally, studies provided limited baseline details and inclusion criteria.

Familial prostate cancer

A total of 11 primary studies (and two related publications) focused on patients with familial PC (Table II). As with the studies targeting unselected populations, the majority included patients from the USA (six studies, including one multinational study) and the UK (two studies). Two studies were identified for Germany (including the single multinational study), and one study each included patients from Australia and Japan. All 11 studies exclusively reported results for patients with unselected PC. For the selected familial PC subgroup, specific DDR mutations are described in Data S1, Appendix S4.

Table II

Study characteristics: Selected familial subgroups.

PC group	Study ID	Other related publications	No. analyzed	Recruitment/enrollment dates	Place(s) of recruitment
A, Australia
PC	Cheng et al, 2011a (88)	NR	147	NR	NR
B, Germany
PC	Maier et al, 2014 (37)	Maier et al, 2010 (59)	382; 92	1998-2007	Ulm, Germany
C, Japan
PC	Hayano et al, 2016 (42)	NR	140	NR	Gunma University Hospital and its affiliated hospitals, Japan
D, UK
PC	Gayther et al, 2000 (89)	NR	38	NR	UK
PC	Leongamornlert et al, 2014 (41)	NR	191	NR	UK Genetic Prostate Cancer Study (UKGPCS)
E, USA
PC	Zuhlke et al, 2012 (90)	NR	94	NR	University of Michigan Prostate Cancer Genetics Project and Johns Hopkins University
PC	LaDuca et al, 2017a (91)	NR	NR	NR	Not clear-patients referred from unknown centers
PC	Ledet et al, 2017a (92)	Lin et al, 2017 (93)	124	2015-2016	Tulane Cancer Center
PC	Nicolas et al, 2015 (39)	NR	12	NR	Fox Chase Cancer Center
PC	Marshall et al, 2017a (94)	NR	92	August 2013- September 2016	NR
F, Multi-national
PC	Hebbring et al, 2006 (85)	NR	1819; 1218	NR	USA (Johns Hopkins, Tampere University Hospital, University of Michigan, Mayo Clinic); Germany (Universitätsklinikum Ulm)

Abstract. NR, not reported; PC, prostate cancer.

DNA damage response gene definition and reported methodology

The DDR genes reported and analyzed are presented in Table III. In total, 40 studies (50%) and 25 studies (31%) analyzed germline and somatic mutations, respectively. Six studies (7.5%) considered both germline and somatic mutations, and nine studies (11%) did not specify the nature of the mutation.

Table III

DDR genes of interest analyzed.

Author, year	Germline or somatic mutation	ATM	ATR	BRCA1	BRCA2	CHEK2	FANCA	MLH1	MRE11A	NBN	PALB2	RAD51C	Other DDR genes	Refs.
A, Australia
Cheng et al, 2011	NR/unclear	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(88)
B, Canada
Nam et al, 2005	Germline	No	No	No	No	Yes	No	No	No	No	No	No	No	(57)
Tischkowitz et al, 2008	Germline	No	No	No	No	Yes	No	No	No	No	No	No	No	(95)
Akbari et al, 2014	Germline	No	No	No	Yes	No	No	No	No	No	No	No	No	(36)
Hamel et al, 2003	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(96)
Struss et al, 2017	Germline	?	?	?	Yes	?	?	?	?	?	Yes	?	?	(23)
Damaraju et al, 2006	Germline	Yes	No	Yes	Yes	No	No	No	Yes	No	No	Yes	Yes	(97)
C, Denmark
Naslund Koch et al, 2016	Germline	No	No	No	No	Yes	No	No	No	No	No	No	No	(58)
D, Germany
Maier et al, 2014	Germline	No	No	No	Yes	No	No	No	No	No	No	No	No	(37)
Meyer et al, 2007	Germline	Yes	No	No	No	No	No	No	No	No	No	No	No	(98)
Nientiedt et al, 2017	Somatic	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(99)
E, Israel
Vazina et al, 2000	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(60)
Hubert et al, 1999	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(100)
Giusti et al, 2003	Somatic	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(101)
F, Japan
Hayano et al, 2016	Germline	Yes	?	?	?	?	Yes	?	?	?	Yes	?	Yes	(42)
Tanaka et al, 2009	Somatic	No	No	No	No	No	No	Yes	No	No	No	No	No	(61)
Uchida et al, 1999	NR/unclear	No	No	Yes	No	No	No	No	No	No	No	No	No	(62)
G, Spain
Romero et al, 2017	Germline	Yes	?	Yes	Yes	Yes	?	?	?	?	Yes	?	Yes	(22)
H, UK
Kote-Jarai et al, 2011	Germline	No	No	No	Yes	No	No	No	No	No	No	No	No	(102)
Edwards et al, 2003	Germline	No	No	No	Yes	No	No	No	No	No	No	No	No	(103)
Leongamornlert et al, 2012	Germline	No	No	Yes	No	No	No	No	No	No	No	No	No	(32)
Castro et al, 2011	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(31)
Gayther et al, 2000	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(89)
Angele et al, 2004	Germline	Yes	No	No	No	No	No	No	No	No	No	No	No	(65)
Leongamornlert et al, 2014	Germline	Yes	No	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	(41)
Jefferies et al, 2017	Somatic	Yes	?	?	?	?	?	?	?	?	?	?	?	(14)
Manson-Bahr et al, 2015	Somatic	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	(27)
I, USA
Agalliu et al, 2007	Germline	No	No	No	Yes	No	No	No	No	No	No	No	No	(104)
Gallagher et al, 2012	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(105)
Agalliu et al, 2009	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(106)
Kirchoff et al, 2004	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(107)
Lehrer et al, 1998	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(108)
Petrovics et al, 2016	Germline	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(46)
Nicolosi et al, 2017	Germline	?	?	Yes	Yes	?	?	?	?	?	?	?	?	(19)
Cesaretti et al, 2007	Germline	Yes	No	No	No	No	No	No	No	No	No	No	No	(109)
Browning et al, 2006	Germline	Yes	No	No	No	No	No	No	No	No	No	No	No	(74)
Zhu et al, 2010	Germline	Yes	No	No	No	No	No	No	No	No	No	No	Yes	(110)
Pomerantz et al, 2017	Germline	Yes	?	?	Yes	?	Yes	?	?	?	?	?	Yes	(111)
Marshall et al, 2017	Germline	Yes	?	Yes	Yes	Yes	?	?	?	?	?	?	Yes	(94)
Ledet et al, 2017	Germline	Yes	?	Yes	Yes	Yes	?	?	?	Yes	?	?	Yes	(92)
LaDuca et al, 2017	Germline	Yes	?	Yes	Yes	Yes	Yes	?	?	?	Yes	?	Yes	(91)
Hart et al, 2016	Germline	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	(112)
Antonarakis et al, 2018	Germline	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	(113)
Nicolas et al, 2015	Germline	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	(39)
Beltran et al, 2015	Germline and somatic (mixed)	No	No	No	No	No	Yes	No	No	No	No	No	No	(79)
Wu et al, 2006	Germline and somatic (mixed)	No	No	No	No	Yes	No	No	No	No	No	No	No	(76)
Abida et al, 2017	Germline and somatic (mixed)	Yes	?	Yes	Yes	Yes	Yes	Yes	?	Yes	Yes	?	Yes	(17)
Zuhlke et al, 2012	NR/unclear	No	No	No	No	No	No	No	No	Yes	No	No	No	(90)
Nastiuk et al, 1999	NR/unclear	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(114)
Daniel et al, 2017	NR/unclear	?	?	Yes	Yes	?	?	?	?	?	?	?	?	(83)
Feldman et al, 2014	NR/unclear	Yes	?	?	?	?	?	?	?	?	?	?	?	(72)
Dawson et al, 2016	NR/unclear	Yes	?	?	Yes	?	?	?	?	?	?	?	Yes	(66)
Dall'Era et al, 2017	NR/unclear	Yes	Yes	Yes	Yes	?	?	Yes	?	?	?	?	Yes	(77)
Williams et al, 1996	Somatic	No	No	Yes	No	No	No	No	No	No	No	No	No	(47)
Myers et al, 2016	Somatic	No	No	Yes	Yes	No	No	No	No	No	No	No	No	(35)
Palapattu et al, 2015	Somatic	?	?	?	Yes	?	?	?	?	?	?	?	?	(73)
Robbins et al, 2011	Somatic	?	?	?	Yes	?	?	?	?	?	?	?	Yes	(84)
Pritchard et al, 2014	Somatic	?	?	?	?	?	?	Yes	?	?	?	?	Yes	(16)
Evans et al, 2016	Somatic	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	(28)
Schweizer et al, 2016	Somatic	?	?	?	Yes	Yes	?	Yes	?	?	?	?	Yes	(115)
Lara et al, 2017	Somatic	?	?	Yes	Yes	?	?	?	?	?	?	?	?	(34)
Grasso et al, 2012	Somatic	Yes	No	No	Yes	No	No	No	No	No	No	No	Yes	(26)
Gourdin et al, 2016	Somatic	Yes	No	Yes	Yes	No	No	No	No	No	No	No	?	(15)
Gambhira et al, 2016	Somatic	Yes	No	Yes	Yes	No	No	Yes	No	No	No	No	Yes	(82)
Patel et al, 2016	Somatic	Yes	?	Yes	Yes	?	?	?	?	?	Yes	?	Yes	(29)
Xia et al, 2015	Somatic	Yes	Yes	?	?	Yes	Yes	Yes	?	Yes	?	?	Yes	(67)
Beltran et al, 2013	Somatic	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	Yes	No	(81)
J, Multi-national
Na et al, 2017	Germline	Yes	No	Yes	Yes	No	No	No	No	No	No	No	No	(25)
Pritchard et al, 2016	Germline	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	(21)
Nelson et al, 2016	Germline	Yes	Yes	Yes	Yes	Yes	?	?	Yes	Yes	Yes	Yes	Yes	(20)
Decker et al, 2016	Germline and somatic (mixed)	?	?	?	Yes	?	?	?	?	?	?	?	Yes	(30)
Robinson et al, 2015	Germline and somatic (mixed)	Yes	?	Yes	Yes	Yes	?	Yes	?	?	Yes	?	Yes	(4)
Lu et al, 2015	Germline and somatic (mixed)	Yes	Yes	Yes	Yes	?	Yes	?	?	?	Yes	?	Yes	(24)
Hebbring et al, 2006	NR/unclear	No	No	No	No	No	No	No	No	Yes	No	No	No	(85)
Fontugne et al, 2015	Somatic	?	?	?	Yes	?	?	?	?	Yes	?	?	Yes	(38)
The Cancer Genome	Somatic	Yes	?	Yes	Yes	?	?	Yes	?	?	?	Yes	Yes	(18)
Atlas, 2015
Timms et al, 2016	Somatic	Yes	Yes	Yes	Yes	Yes	Yes	?	?	Yes	Yes	Yes	Yes	(13)
K, NR
Stephens et al, 2016	Somatic	?	?	?	Yes	?	?	?	?	?	?	?	?	(116)
Sonpavde et al, 2017	Somatic	?	?	Yes	Yes	?	?	?	?	?	?	?	?	(33)
Liu et al, 2016	Somatic	Yes	?	?	?	?	?	Yes	?	?	?	?	?	(87)

?, unclear; ATM, ataxia telangiectasia mutated; ATR, ataxia telangiectasia and Rad3-related protein; BRCA, breast cancer susceptibility gene; CHEK2, checkpoint kinase 2; DDR, DNA damage repair; FANCA, Fanconi anemia complementation group A; MLH1, MutL homolog 1, MRE11A, MRE11 homolog A, double-strand break repair nuclease; NBN, nibrin; NR, not reported; PALB2, partner and localizer of BRCA2; RAD51C, RAD51 paralog C.

The source of DNA and the methods of genetic analysis varied considerably among studies (Data S1, Appendix S6). The major sources of DNA were blood (39 studies, 48.1%) and tumor tissue (28 studies, 34.6%). Six samples were sourced from cell-free DNA (7.4%), five from paraffin-embedded tumor tissue (6.2%), and four from buccal swabs/saliva (4.9%). One study sourced DNA from a single-cell suspension of a primary tumor (1.2%). Nine studies did not report the source of the DNA analyzed (11.1%).

The two most frequently used methods of mutational analysis were PCR (29 studies, 35.8%) and next-generation sequencing (19 studies, 23.5%). Seven studies performed whole-exome sequencing (8.6%), three used Sanger sequencing (3.7%), three used capture sequencing (3.7%) and two used fluorescence in situ hybridization (FISH; 2.5%). A total of 10 studies used methods other than those specified within this report (12.3%), whereas six studies (7.4%) did not report which methods were used.

DDR genes could be separated into three general categories: Those that focused on specific mutations in a specific gene (21 studies, 25.9%); those that focused on all, or undefined, mutations in a specific gene (22 studies, 27.1%); and those that focused on undefined mutations in multiple DDR genes (38 studies, 46.9%). Definitions for individual gene mutations or DDR combinations varied considerably from study to study.

Risk of bias of included studies

None of the studies were free from the risk of bias based on the Joanna Briggs Institute Critical Appraisal Checklist; 88% (of the 80 included studies) had at least one domain judged to be high risk (Data S1, Appendix S7 and Fig. S1). The majority of studies had a sample size of >50 participants (79% of studies), an analysis with sufficient coverage/clearly described mutations (84%), and appropriate statistical analysis with a clear prevalence calculation (74%). Overall, however, the quality of reporting was poor and did not allow for judgment of how well participants were recruited or how data were acquired; consequently, 76% of studies were judged to have an unclear risk of bias. Likewise, the criteria for the diagnosis of PC (e.g. the Gleason score) were unreported in the majority of studies (74%), and it was unclear in most studies (91%) whether a pathologist had performed the diagnosis. Confounding factors were not taken into consideration for prevalence reporting in the majority of studies (71%). In addition, the majority of studies (73%) did not describe one or more of the participant baseline characteristics for this review.

Results for prevalence in the unselected population

All studies excluded at the analysis stage are summarized in Data S1, Appendix S3, together with prevalence calculations. To establish if there were patterns of prevalence between somatic and germline mutations, between different genes, and between different PC types, all median prevalence values for all gene datasets were compared (Fig. 2 and Data S1, Appendix S8). Patients identified as having primary PC were grouped under unselected PC; thus, unselected PC was a mixed group in terms of overall diagnosis.

Figure 2

Summary of median prevalence for germline (blue) and somatic (green) DDR gene mutations in unselected populations. Study numbers, median prevalence, range and original data are reported in Data S1, Appendix S8. Data from studies with >50 patients are included. ATM, ataxia telangiectasia mutated; ATR, ataxia telangiectasia and Rad3-related protein; BRCA, breast cancer susceptibility gene; CHEK2, checkpoint kinase 2; DDR, DNA damage repair; FANCA, Fanconi anemia complementation group A; MLH1, MutL homolog 1; MRE11A, MRE11 homolog A, double-strand break repair nuclease; NBN, nibrin; NR, not reported; PALB2, partner and localizer of BRCA2; RAD51C, RAD51 paralog C; PC, prostate cancer; mPC, metastatic prostate cancer; mCRPC, metastatic castration-resistant prostate cancer.

The most evidence was available for DDR gene panels [14 datasets; 11 studies (13-23)], ATM [17 datasets; 13 studies (13,17,18,20-22,24-29)], BRCA1 [19 datasets; 14 studies (17,18, 20-22,24,25,28,30-35)] and BRCA2 [27 datasets; 19 studies (17,18,20-26,28,30,31,33-38)]. Other genes had between four and 11 datasets. For individual genes, the most common mutations were in ATM [median, 1.5-6% across prostate groups (general PC, mPC and mCRPC); full reported range, 0-12%], BRCA2 (median, 1.1-5.2% across prostate groups; full reported range, 0-11.8%), and PALB2 (median, 0-4% across prostate groups; full reported range, 0-4%). The median prevalence of all MutL homolog 1 (MLH1), BRCA1 and Fanconi anemia complementation group A (FANCA) mutations was <4%, and the median prevalence of all checkpoint kinase 2 (CHEK2), NBN, RAD51 paralog C (RAD51C), ataxia telangiectasia and Rad3-related protein (ATR) and MRE11 homolog A, double-strand break repair nuclease (MRE11A) mutations was <2%.

In unselected PC populations and datasets including >500 patients, the frequency of somatic mutations was higher than that of germline mutations in ATM (3.9% vs. 1.5%, respectively), ATR (0.6% vs. 0%), BRCA1 (1.1% vs. 0.6%), BRCA2 (4.9% vs. 1.1%), MLH1 (0.6% vs. 0%), NBN (1.2% vs. 0.3%), PALB2 (1.3% vs. 0.5%) and RAD51C (1.5% vs. 0.5%). By contrast, germline mutations were more common than somatic mutations in CHEK2 (1.8% vs. 1.2%, respectively) and MRE11A (0.2% vs. 0%). In mCRPC, somatic mutations were more common than germline mutations for BRCA1 (2.8% vs. 0.8%, respectively), whereas somatic and germline mutations were equally common for BRCA2 (5% for each group). The prevalence rates for other prostate subgroups were based on <500 patients and were not summarized.

Germline mutations were more common in patients with mPC (based on datasets that included >500 patients) than in patients with general PC for ATM (1.7% vs. 1.5%, respectively), ATR (0.3% vs. 0%), BRCA1 (0.9% vs. 0.6%) and BRCA2 (5.2% vs. 1.1%), but not for CHEK2 (1.5% vs. 1.8%), NBN (0.2% vs. 0.3%), PALB2 (0.5% for both groups) or RAD51c (0.2% vs. 0.5%). Germline mutations were more common in mCRPC than in general PC for BRCA1 (0.8% vs. 0.6%) and BRCA2 (5.0% vs. 1.1%). Fewer than 500 patients were available for other prostate subgroups, and prevalence in those groups was therefore not summarized.

Somatic mutations were more common in the mCRPC population (based on datasets including >500 patients) than in the unselected PC population for BRCA1 (2.8% vs. 1.1%), whereas similar rates of somatic mutations were seen in the two populations for BRCA2 (5.0% vs. 4.9%).

The prevalence for DDR genes as a combined term was much higher than for the individual genes. This was expected, as the definition was based on the presence of multiple gene mutations (Table III and Data S1, Appendix S8). In general PC, the prevalence of somatic DDR gene mutations was in the range of 4.9-22%, while germline DDR mutation rates ranged between 17.2 and 19%. In mPC, the prevalence of somatic DDR mutations ranged between 10 and 16.4%, while germline mutation rates ranged between 11.4 and 11.8%. In mCRPC, the prevalence of germline DDR mutations was in the range of 7.5-9.1%; no somatic mutations were reported. In general PC, based on datasets including >500 patients, germline mutations had a higher prevalence than somatic mutations (18.6% vs. 10.7%, respectively); other somatic subgroups included <500 patients, and so prevalence rates were not analyzed. Germline DDR mutation rates were higher in patients with general PC (18.6%) than in those with mPC (11.6%) or mCRPC (8.3%).

Two multinational studies (4,30) and one US study (17) reported the prevalence of mutations using definitions that combined germline and somatic variants (Table IV). These studies used different sequencing methods to investigate mutation rates in mCRPC (4,30) and general PC (17). Decker et al (30) reported the prevalence for patients who had both somatic and germline mutations, whereas Robinson et al (4) and Abida et al (17) reported the prevalence for patients who had either a somatic or a germline mutation (or both). Thus, the rates in the study of Robinson et al (4) for BRCA1 (2.67%) and BRCA2 (12.67%) are approximately double those in the study of Decker et al (30) (0.66 and 6%, respectively). The combined prevalence of germline and/or somatic DDR gene mutations was 22.67% in mCRPC (4) and 27% in general PC (17).

Table IV

Summary of DDR gene mutation prevalence in combined definitions for germline and/or somatic tissues.

Author, year	Study information	Definition of combined germline and somatic	Method of mutation identification	Inclusion criteria	Gene	Gene definition	No. of patients	% prevalence	Refs.
Decker et al, 2016	mCRPC; Multi-national	Patients who had both germline and somatic mutations	Whole- genome sequencing	Discovery set PC patients were selected based on high Gleason score and availability of both peripheral blood DNA and fresh frozen prostatectomy samples. 150 samples from Robinson 2015 (4).	BRCA2	Biallelic mutation	150	6	(30)
					BRCA1	Ballelic loss	150	0.66
Abida et al, 2017	PC; USA	Patients who had either germline or somatic or both mutations	Next- generation sequencing	Mixed population of locoregional (n=50), biochemically recurrent (n=53), and metastatic (n=348). 221 samples were included of unclear pathology, including 124 samples from Pritchard 2016 (21).	DDR	Alterations of BRCA2, BRCA1, ATM, CHEK2	221	27	(17)
Robinson et al, 2015	mCRPC; Multi-national	Patients who had either germline or somatic or both mutations	Whole-exome sequencing	Affected individuals who had metastatic disease accessible by image-guided biopsy and were being considered for abiraterone acetate or enzalutamide as standard of care or as part of a clinical trial were considered for enrollment	ATM	ATM mutations (total)	150	5.33	(4)
					BRCA2	Loss of BRCA2	150	12.67
					BRCA1	BRCA1 mutations (total)	150	2.67
					CHEK2	CHEK2 mutations (total)	150	3.33
					FANCA	FANCA mutations (total)	150	0.67
					MLH1	MLH1 mutations (total)	150	2.00
					PALB2	PALB2 mutations (total)	150	2.00
					DDR	Aberrations in DNA repair/recombination	150	22.67

All studies were in unselected populations from multi-national countries. ATM, ataxia telangiectasia mutated; BRCA, breast cancer susceptibility gene; CHEK2, checkpoint kinase 2; DDR, DNA damage repair; FANCA, Fanconi anemia complementation group A; mCRPC, metastatic castration resistant prostate cancer; MLH1, MutL homolog 1; PALB2, partner and localizer of BRCA2; PC, prostate cancer.

Results for prevalence in the familial subgroup

The prevalence of BRCA2 and DDR mutations in familial PC is summarized in Fig. 3 and Data S1, Appendix S4. The median prevalence of germline BRCA2 mutations in the familial PC subgroup was 3.7% (range, 1.3-7.9%), based on six studies (n=945). While these studies reported variable definitions for familial PC, they generally required more than two family members with PC. Only one-half of the studies reported baseline characteristics (Data S1, Appendix S5); therefore, further analysis of the influence of baseline details was not possible. The median prevalence of germline BRCA2 mutations (3.7%) was notably higher than that in the unselected population (1.1%).

Figure 3

Summary of median prevalences for DDR and BRCA2 gene mutations in patients with familial prostate cancer. 'All BRCA2' and 'All DDR' signify all mutations identified for each gene, whereas 'specific BRCA2' signifies one specific mutation (e.g. a specific deletion) in the gene. Study numbers and BRCA2/DDR definitions are reported in Data S1, Appendix S4. BRCA, breast cancer susceptibility gene; DDR, DNA damage repair.

The median prevalence of germline DDR mutations in familial PC was 29.3% (range, 7.3-91.67%), based on three studies (n=327). It is noteworthy that one other study (39) provided two definitions for 'DNA damage response or androgen-signaling gene variants' (Data S1, Appendix S6); one definition referred to any affected gene (91.7%), whereas the other referred to two or more affected genes (41.7%). The inclusion of 'androgen-signaling gene variants' in the definition of DDR was inconsistent with the definitions of DDR in the other studies, and therefore this study was excluded from a sensitivity analysis, leading to a median prevalence of 12.1% (range, 7.3-16.9%).

Discussion

To the best of our knowledge, this is the first review that has used rigorous systematic review methods (8,9) to report a comprehensive summary of recently published data on the prevalence of DDR genes in PC (including in mPC, mCRPC and selected subgroups).

The most common mutations (measured by median prevalence) in all unselected populations were ATM, BRCA2 and PALB2. The highest median reported rates for germline mutations in BRCA2 were found in mPC and mCRPC. The highest median reported rates for somatic mutations in mCRPC were found for ATM and BRCA2.

Overall, the median prevalence for DDR germline mutations was 18.6% in general PC (range, 17.2-19%; n=1,712), 11.6% in mPC (range, 11.4-11.8%; n=1,261), and 8.3% in mCRPC (range, 7.5-9.1%; n=738). The median prevalence for DDR somatic mutations was 10.7% in general PC (range, 4.9-22%; n=1,537) and 13.2% in mPC (range, 10-16.4%; n=105).

The prevalence for DDR germline and/or somatic mutations was 27% in general PC (17) and 22.67% in mCRPC (4). The higher rate of 27% (17) for germline and/or somatic mutations in general PC may reflect the mixed population of the sample, which was predominantly metastatic rather than localized; it may also reflect differences in the genes included in the definition of 'DDR.' A recent study by Armenia et al (40) (after the search dates of this review) reported a combined germline and somatic mutation rate of 27% for DDR genes in mCRPC. This rate is similar to those reported here. Differences in the precise rates for mCRPC reported by Armenia et al (40) and Robinson et al (4) could be due to variations in the genes included in the definition of 'DDR' or the use of different sequencing methods (next-generation sequencing vs. whole-exome sequencing). The median rate of germline BRCA2 mutations in patients with familial PC history was 3.7%, higher than that in the equivalent unselected population (1.1%). Studies report that DDR pathways are good candidates for PC predisposition and that, although BRCA2 is the most frequently mutated gene, a wider range of DDR genes are likely to predispose to PC (41,42).

These results should be interpreted with caution given the limited reporting of baseline characteristics and the heterogeneity of sequencing methods and mutational definitions. In particular, the patients included in the 'unselected PC' population varied widely. For example, unselected PC cases in Abida et al (17) represented a mixed population of cancer types (local and metastatic) dominated by metastatic samples (77%), whereas patients from the Cancer Genome Atlas (18) were described as primary since they were derived from prostatectomies, but in fact included many high-grade cancers that were likely to be metastatic.

No other systematic reviews on the prevalence of DNA repair gene mutations and PC were identified. Two recent papers by Isaacsson Velho et al (43) and Quigley et al (44) identified similar mutational rates to those reported in this review. Isaacsson Velho et al (43) found a germline mutational rate of BRCA2 in mPC of 6%, compared with the median prevalence of 5.2% reported here. Quigley et al (44) found germline and/or somatic BRCA2 rates in mCRPC of 10%, compared with the rate of 12.7% (4) reported here. It is likely that the differences in rates may be explained by differences in sequencing methodology, precise definitions of mutations, and/or differences in populations. Another recent paper (45) used The Cancer Genome Atlas (18) tissues and those reported by Armenia et al (40) to investigate the influence of Gleason score and tumor stage. The overall prevalence of somatic DDR gene mutations in localized tumors was 8%, which is within the range reported here (4.9-22%). Marshall et al (45) identified an increase in DDR mutation prevalence in patients with Gleason grade ≥3 and clinical stage ≥cT3 disease.

The findings of the present review provide evidence in support of the testing of DDR germline mutations in advanced disease, in line with NCCN and Philadelphia Prostate Cancer Consensus recommendations, and the St Gallen Advanced Prostate Cancer Consensus Conference 2017 (5-7). Some germline DDR mutations may present an increased healthcare burden, since family members may also be at risk of PC. Depending on which DDR gene mutations are present, family members may also have an increased risk of breast, ovarian, pancreatic and colorectal cancer, melanoma and other cancer types (5). Given that somatic mutations were found to have a higher or similar prevalence compared with germline mutations in patients with mCRPC, somatic mutations may provide useful genetic information to guide participation in clinical trials or additional mutation testing. These findings support the testing of all patients with metastatic disease and not just those with familial disease. BRCA2 was identified as having the highest mutation rate of any individual DDR gene, supporting the use BRCA2 screening as recommended across all consensus conferences and NCCN (5-7).

Of note, some of the included studies provided limited baseline details. Where reported, there was evident heterogeneity in terms of the period of data collection, data sources, previous treatments, sequencing/screening methods and the risk of bias. Prevalence data based on low patient/study numbers combined with limited information about and/or variation in study characteristics prevented a thorough interrogation of the results, and the reasons for variation in prevalence could not always be determined. The repository data in many of the studies included in this review were unaccompanied by baseline patient details (13), and the source of the samples was often mentioned only incidentally within the results. This hampered data extraction and increased the risk of double counting (using the same data source twice). For example, among the three sets of tissues analyzed in the 2016 report of Decker et al (30), one was the same tumors used in the 2015 study of Robinson et al (4). Despite this apparent overlap, the data presented in the two papers do not double count for any result presented. To avoid double counting, only original data were extracted from each paper. Study sizes were limited, and only eight studies included >500 participants (20,21,25,31,32,34,36,46).

Definitions for individual gene mutations or DDR combinations varied considerably from study to study. Sources of tissue varied (being either fresh or paraffin-embedded), as did methods of mutational analysis. Different DNA sequencing/screening methods used for the detection of DDR mutation may explain the variations observed in the prevalences reported; this was demonstrated by the two linked publications of Williams et al (47) and Gao et al (48). Both publications used the same 23 patients with PC, but reported using different techniques to identify BRCA1 mutations. The first (48) detected loss of heterozygosity, using PCR, at the BRCA1 locus in 22% of samples, whereas the second paper (47), using FISH, found that the loss was not in the BRCA1 gene but was distal, and therefore reported a 0% prevalence. Small gene panels and circulating DNA analysis may have lower coverage of large tumor suppressor genes such as ATM, BRCA1 and BRCA2 compared to whole genome or whole exome sequencing.

Differences in the definitions of what is considered a mutation can also arise, such as the use of monoallelic versus biallelic approaches to categorizing DDR mutations compared with non-mutations. One study (49) reported a comparison between DDR mutations (defined as biallelic mutations in DDR genes) compared with non-mutations (defined as wild-type or monoallelic mutations in DDR genes, which the authors considered non-deleterious). By contrast, all other studies (with one partial exception) reported any mutation in a DDR gene (whether it affected one or both alleles) as a bona fide DDR mutation. The exception involved a study (13) that provided two definitions for DDR, alternately based on one or both alleles being defective.

The use of different reference genome builds as comparator sequences introduces uncertainty. It is standard practice to use such builds as normal comparator sequences in order to identify mutations, but these reference sequences may change over time (with improved methodologies), which may influence what is defined as a genuine mutation. The use of different sequencing depths introduces uncertainty for between-study comparisons. For example, studies that perform next-generation sequencing to an extended depth [e.g. x265 (38)] may achieve more accurate sequencing (lower error rates) than studies that sequence to a relatively minimal depth [e.g. x100 (27)]. The use of different frequency cut-off thresholds for the inclusion of mutations introduces uncertainty for between-study comparisons. For example, studies that apply a relatively high frequency threshold when analyzing exonic mutations (e.g. excluding mutations with frequencies >2% in the observed population) (37) will include a larger pool of mutations than studies that apply a more restrictive frequency cutoff (e.g. excluding mutations with a population allele frequency ≥0.5%) (50).

There is a risk of sampling bias with somatic mutations, given that PC is a very heterogeneous disease (51). This heterogeneity means that within a patient tumor there are multiple sub-tumors and normal tissues with different genotypes (different mutational profiles) or clones (52,53). For example, PC can have areas of moderately differentiated tumor and undifferentiated tumor, and each clone will have a different prognosis. Tumor heterogeneity impacts biomarker studies, since the sample taken for analysis may not be the same as the sample given to the pathologist, and false associations between biomarker and histological stage can arise. Any method other than micro-dissected tissue confirmed by a pathologist for tumor grade is at risk of sampling error.

In the present study, no prevalence data were available for CRPC populations with >50 participants. There was insufficient evidence to judge whether patterns of prevalence were influenced by the country of recruitment. While there was some evidence for all the DDR genes on which the study focused, evidence was particularly limited for ATR, FANCA, MLH1 and MRE11A, as mutations in these genes were reported in five or fewer studies in total.

Future research should focus on other subgroups (including African-Americans, and young-onset, ductal and lethal prostate cancer) that may have DDR mutation prevalences that are different from that of the wider PC population. In addition, future work should consider the contribution of founder mutations in unselected populations and a consideration of mutations that are prevalent in other countries. A more focused review would be required to examine in detail whether mutations were pathogenic or were variants of uncertain significance.

In future research, authors should better clarify patient baseline characteristics, the diagnostic methods employed, and whether the somatic samples used for analysis were diagnosed directly or inferred from other pathology samples. Assessment of data according to cancer stage and grade rather than broad terms such as 'PC' may also be more useful. Definitions of what types of genetic changes constitute 'mutations' (e.g. single-copy deletion compared with homozygous deletion) need to be standardized, and pathogenic classification should be routinely incorporated into the reporting of mutations. Finally, reviews of biomarker studies need to consider the aforementioned scientific limitations of these studies, as well as study design limitations. Future guidelines/consensus are required for greater standardization of sequencing methods used for genetic association studies.

Given the generally poor quality and poor reporting of studies showing prevalence data, there is a need for future epidemiological studies that use recognized methodologies and reporting tools. Authors should be encouraged to adhere to the following reporting guidelines for prevalence studies: The Simon et al (54) guidelines for the use of archived specimens in the evaluation of prognostic and predictive biomarkers; Biospecimen Reporting for Improved Study Quality (55); Strengthening the Reporting of Genetic Association Studies, an extension of the STROBE statement (56); and the Joanna Briggs Institute Critical Appraisal Checklist for Studies Reporting Prevalence Data (to reduce the risk of bias) (10).

A number of recent updates (5-7) in clinical practice guidelines have followed the arrival of new types of genetic tests and ongoing clinical trials with drugs that specifically target germline and/or somatic DDR mutations in mCRPC. In the unselected PC population, the median prevalence of germline and/or somatic DDR mutations was 27%. In the mPC populations, the median rate of DDR mutations was 11.6% for germline mutations and 13.2% for somatic mutations. In mCRPC populations, the prevalence rate was 22.67% for germline and/or somatic mutations. In patients with a familial history of PC, the median rate of germline DDR mutations was 12.1% and the median rate of germline BRCA2 mutations was 3.7%.

The present review highlighted variations in definitions and methodology among studies investigating biomarkers, including differences in tumor sampling, tumor heterogeneity, sequencing depth and mutational frequency thresholds, and comparability of mutational definitions (genetic terminology). Further large, well-reported prevalence studies in PC are needed to provide international prevalence data on the burden of DDR dysfunction in patients with PC.

Funding

The present study was sponsored by Pfizer, Inc.

Availability of data and materials

All data analyzed during this study are included in this published article.

Authors' contributions

SHL, SLS, JK, RGWQ contributed to the conception and design, acquisition of data, analysis and interpretation of data. HW contributed to the acquisition of data, analysis and interpretation of data. KM designed and performed the search strategies. All authors contributed to the writing of the manuscript and its final approval.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

SHL, SLS, HW, KM and JK are employees of Kleijnen Systematic Reviews Ltd., who were paid consultants to Pfizer, Inc. in connection with the development of this manuscript; they have no other competing interests. RGWQ is an employee of Pfizer, Inc..