Randomized trial finds Prostate Cancer Genetic Risk Score Feedback Targets PSA Screening Among At-Risk Men

Abstract

Background

Prostate-specific antigen (PSA) screening may reduce death from prostate cancer (PC), but leads to overdiagnosis of many cases of indolent cancer. Targeted use of PSA screening may reduce overdiagnosis. Multi-marker genomic testing shows promise for risk assessment, and could be used to target PSA screening.

Methods

To test whether counseling based on family history (FH) versus counseling based on Genetic Risk Score (GRS)+FH differentially affects subsequent Prostate Specific Antigen (PSA) screening at 3 months (primary outcome), we conducted a randomized trial of FH vs. GRS+FH was conducted in 700 Caucasians aged 40-49 years without prior PSA screening. Secondary outcomes included anxiety, recall, physician discussion at three months, and PSA screening at 3 years. We also evaluated pictographs versus numeric presentations of genetic risk.

Results

At three months, no significant differences were observed in rates of PSA screening between FH (2.1%) and GRS+FH (4.5%) arms (x²=3.13, p=0.077); however, PSA screening rates at three months significantly increased with given risk in the GRS+FH arm (p=0.013). Similar results were observed for discussion with physician at three months and PSA screening at three years. Average anxiety levels decreased after providing individual cancer risk (p=0.0007), with no differences between groups. Visual presentation by pictographs did not significantly alter comprehension or anxiety.

Conclusion

This is likely the first randomized trial of multi-marker genomic testing to report genomic-targeting of cancer screening. We found little evidence of concerns for excess anxiety or over/under use of PSA screening when providing multi-marker genetic risks to patients.

Graphical abstract

Precis: This prospective trial found that provision of individual genetic risk scores for prostate cancer did not lead to significant increases in anxiety or use of PSA screening. Rather, genetic risk scores led to targeted use of PSA screening among higher risk men, which may improve PSA screening performance.

Introduction

The United States Preventive Services Task Force recommends against routine Prostate-Specific Antigen (PSA) screening, citing risks of over-diagnosis and over-treatment¹. The American Cancer Society (ACS), American Urological Association (AUA), and European Association of Urology (EAU) maintain that PSA screening should be considered after physician-patient discussion, and earlier age screening (<age 55 years) should be offered to men at increased risk (e.g. positive family history)^{2, 3, 4}. Prospective trials suggest that a favorable balance of risks and benefits can be achieved by targeting PSA screening to higher risk patients.^{5, 6, 7}

Family history (FH) risk assessment is an established approach that enables targeted PCa screening in clinical settings, and is actively promoted by The Centers for Disease Control and The US Surgeon General.^{8, 9} Genetic Risk Score (GRS) is a new approach to determine individual risk that builds on family history by assessing single nucleotide polymorphisms (SNPs) associated with PCa risk.^10-14 The clinical validity and utility of risk prediction using GRS has been demonstrated in large prospective studies.^15-16 However, there are concerns for clinical application of GRS, including increased anxiety and avoidance or overuse of medical care.^17-20 In consideration of these concerns, we studied a specific application of GRS plus FH, for personalized risk assessment of PCa, and evaluated the potential for targeted Prostate Specific Antigen (PSA) screening.

Methods

Design

We conducted a prospective, randomized-controlled study primarily comparing the impact of GRS+FH versus FH risk feedback information on PSA screening rates as the primary outcome. Secondary outcomes included discussion of PSA screening with a physician and subject anxiety. We also compared anxiety and PCa risk recall for numeric versus pictograph presentations of this risk feedback for a 2×2 matrix of 4 randomization groups. Prior studies suggested pictographs increase comprehension and reduce worry.²¹ The design and data collection timepoints are shown in Figure 1.

Figure 1. Study design, enrollment, and outcomes.

Randomization

Computerized randomization of 700 study IDs into 175 blocks of four each was completed prior to study enrollment, ensuring subjects were continuously randomized to each of the four randomization groups during the recruitment timeframe. The randomization list was kept at Wake Forest, and not available to the enrollment team in Michigan. This process masked randomization status until the risk report was provided to each participant. Randomization groups were (Group 1) GRS+FH as a number; (Group 2) GRS+FH as a number and pictograph; (Group 3) FH as a number; and (Group 4) FH as a number and pictograph.

Enrollment

Recruiters worked with primary care offices in West Michigan (Spectrum Health Medical Group and Grand Valley Medical Specialists) to query patient databases for qualified patients from June 2011 through February 2012. Potential participants were screened over the phone or in person and qualified individuals received detailed information about undergoing genetic testing for PCa and randomization details as part of the study informed consent process. Each subject was offered personalized risk feedback and up to $80 compensation if they completed the trial.

Eligibility criteria were age 40 to 49 years, self-defined Caucasian background, and no prior PSA screening nor PCa diagnosis. These criteria helped ensure inclusion of high-risk men that were PSA-naïve, and were consistent with ACS and AUA guidelines for earlier offer of screening for men at increased risk. Participation was limited to Caucasians due to lacking information regarding risk prediction with these genetic markers in other races/ethnicities at the time of study startup. Informed consent was documented in writing. Prior to enrollment, this study was approved by institutional review boards at each participating institution.

Data collection and study participation

At baseline, participants completed the State-Trait Anxiety Inventory and donated saliva samples (Oragene, Inc). A CLIA-certified lab at Spectrum Health Hospital in Grand Rapids Michigan assayed a validated panel of 46 SNPs. A report for lifetime risk of PCa was generated for each subject (Supplemental Figure 1). The report content and format was based on the 4 randomization groups. Participants in the FH arm were given lifetime PCa risks of either 17% or 23% based on negative or positive family history (Figure 2a), respectively, based on the estimated OR for FH, calibrated incidence rates, and mortality rates excluding PCa as derived from SEERS data.^{12, 22} In the GRS+FH arm, risk calculations included the SEERS-derived FH+/− risk and a sum of the estimated OR for each SNP, resulting in risks that were distributed more broadly (Figure 2b) based on the number of risk alleles.^{12, 22} Four to six weeks after baseline, participants met with a certified genetic counselor to receive their personal risk report and complete survey measures. At the three-month phone survey, participants repeated survey measures, with new questions about behaviors following the intervention, including further discussion of PCa screening with their PCP and PSA screening. Electronic medical records of all subjects were queried for PSA screening (median: 3.02 years).

Figure 2. Distribution of Genetic Risk Scores Given to Study Participants in FH (A) and GRS+FH (B) arms.

Red bars represent the percentage of participants given a specific risk value. Black dashed line at 18.3% represents the average risk that was also provided on each participant risk report.

Genetic Testing & Risk Reports

All participants were asked to donate saliva samples (Oragene, Inc) to maintain the blind status of participants until the results were disclosed. All saliva samples and data were labeled with unique study identifiers to protect confidentiality. Samples were transported to a CLIA-certified laboratory at Spectrum Health for DNA isolation and SNP genotyping. The lab assayed a validated panel of 46 SNPs, consisting of 33 analytical SNPs and 13 built-in quality controls and duplicated SNPs. The design of this panel was based on prior published research by our study team, with SNPs drawn from the results of prior GWAS and confirmation studies in Caucasian populations (p-value <1.0×10⁻⁶) and limited to one SNP from each independent LD block.^12,22 The Illumina BeadXpress Reader was selected for genotyping based on several factors, namely the 96 well format that allowed a total of 48 samples and controls to be processed in duplicate, ability to multiplex (simultaneous genotype) the 46 selected SNPs per sample, and FDA clearance for in vitro diagnostics. De-identified SNP genotyping results were then sent from the lab at Spectrum Health to the Wake Forest team for risk report generation.

The Wake Forest team subsequently generated a de-identified risk report for each subject, containing the estimated lifetime risk for PCa and a description of factors comprising this risk (Supplemental Figure 1). The content and format of these reports were based on the 4 randomization groups. Pictographs were included in half of the risk reports and were based on a pictograph generator from the University of Michigan.²³ Participants who responded “no” or “do not know” for FH of prostate cancer were classified as having a negative family history. Absolute risk was calculated using the method we have described previously,^12,22 which is based on the SNP-specific RR, calibrated incidence rate of PCa, and mortality rate for all causes excluding PCa in the U.S. Risks greater than 80% were reported as 80% due to concerns about stability of estimates above that point. The risk reports were then transmitted via a secure study website to the study team in Michigan in preparation for visit two. The study coordinator in Michigan linked the study ID back to personal identifiers, added these identifiers, and then finalized the re-identified risk report.

Genetic counseling

Genetic counselors followed prepared scripts when disclosing results (Supplemental methods 1). All deviations/unexpected questions were recorded. Each subject was given a copy of a Centers for Disease Control brochure on PCa screening, and provided with a resource card for more information (Supplemental Figures 4 and 5).

Measures

The State-Trait Anxiety Inventory (STAI) is a validated assessment tool that has been used extensively in both clinical and research settings to measure anxiety.²⁴ It comprises separate self-report scales for measuring state and trait anxiety. In order to measure changes in anxiety at three study timepoints (visits 1, 2-pre, and 2-post; figure 1), we utilized the S-Anxiety scale (State Anxiety). The S-Anxiety scale assesses current feelings “at this moment”: 1) not at all, 2) somewhat, 3) moderately so, and 4) very much so. While the full S-Anxiety scale includes 20 items, we utilized a shortened version, consisting of ten items (Questions 1, 3, 5, 9, 11, 12, 13, 15, 17, and 19) from the STAI Form X1. We referred to published item-remainder correlations that are specific to adult males age 40-49 years, when selecting the subset of items used in this shortened STAI.²⁴ For males age 40-49, the median item-remainder correlation of the full STAI S-anxiety scale is 64, compared to a median of 66 for the ten items selected for the shortened scale used in our study, supporting the use of this subset of items for this study population.²⁴ Each item within the STAI is scored on a scale of 1 to 4, and with ten items the possible range of total scores would be from 10 (lowest anxiety) to 40 (highest anxiety) for each participant. To identify anxiety levels of potential clinical importance, we utilized a pro-rated S-Anxiety score threshold equal to 41 or greater (i.e. threshold=20.5 on scale of 10-40); this threshold was selected based on gender and age-specific normative data in the STAI manual, and then setting a cutoff 0.5 S.D. above the respective mean^24,25. The STAI was assessed at baseline, immediately pre-result, immediately post-result, and at 3 month follow-up.

Risk recall was assessed immediately post result, by the question:

• “Based on the information given to you, what were you told is your chance of developing prostate cancer in your lifetime from 0-100%?” [fill in the blank]

At the 3 month follow-up, behavioral outcomes were assessed by the following questions:

• “Since we last spoke to you, have you talked to a doctor about having a PSA test to further determine your chance of having prostate cancer?” [yes or no]

• “Since we last spoke to you, did you have a PSA test performed?” [yes or no]

Statistical analyses

The primary outcome was self-reported PSA screening by 3 months. Secondary outcomes included: 1) State-Trait Anxiety Inventory (immediate pre/post), 2) risk recall (immediate post, 3 months), 3) discussion with a physician regarding PSA screening by 3 months, and 4) medical record confirmed PSA screening by 3 years. We assessed main effects for risk type [GRS+FH (Groups 1 and 2) versus FH (Groups 3 and 4)]. Non-parametric one-way ANOVA was used to compare continuous data across groups. Fisher’s exact test or chi-square test were used to evaluate associations between categorical measures and groups. Associations between risk estimates provided to subjects and continuous or binary outcomes were tested using linear (recall) or logistic (physician discussion, and had PSA screening) regression, respectively. Repeated measures (anxiety) were evaluated using Wilcoxon signed-rank tests. Statistical analyses were performed using SAS 9.2.

Results

Sample Characteristics

700 patients were enrolled and 97.4% completed the 3-month follow-up. During 3 years of follow-up by medical records, 70.3% were positively observed to receive any health care subsequent to study participation, such as clinic visits or labwork (Figure 1). The demographic distributions suggest randomization resulted in four groups with similar characteristics (Table 1). The average participant was age 45 years, had annual income of $50,000 to $100,000 (46%), a college graduate (45%), married (76%), and a negative first degree FH (91.2%, including 14% in whom FH was unknown).

Table 1.

Subject demographics by randomization group.

	RANDOMIZATION GROUP				p-value*
	GRS-number	GRS-pict	FH-number	FH-pict
Participants (N)
Enrolled	175	175	175	175
Completed visit 2	175	173	175	172	0.62
Completed 3 mo. followup	172	168	173	169	0.24
Medical records observation during 3 yrs. follow-up	125	134	115	118	0.12
Age in years (mean, SD)	44.9 (2.85)	44.8 (3.06)	44.7 (2.84)	45.0 (2.76)	0.76
Annual Income in U.S.D (%)					0.2
<20K	2	4	4	3
20-50K	15	22	22	19
50-100K	47	38	45	47
>100K	33	29	28	29
declined	3	7	2	2
Education (%)					0.54
<12th grade	0	0.6	0	0.6
High School Grad.	8	9	11	10
Some College/Trade	21	29	26	29
College Grad.	49	45	45	39
Post-grad.	21	17	19	21
% declined	0.6	0.6	0	0
Marital Status (%)					0.77
married	77	77	73	75
single	9	10	13	11
divorced	13	10	13	12
widowed	0	0	0	0
Positive Family history PCa (%)	7	14	6	10	0.062

^*t-tests for continuous data, and Fisher’s exact tests for categorical data.

PSA Discussions and Screening by Group

Three months after risk counseling, 16.3% (n=111 of 681) of participants reported discussion of PSA screening with a physician, and 3.25% (n=22 of 677) reported having PSA screening (Supplemental Table 1). For the primary outcome, no significant differences were observed in the rate of PSA screening at three months across by type of feedback (GRS+FH versus FH, x2=3.13, p=0.077); similar non-significant results were seen when comparing across all four randomization groups (x²=3.25, p=0.35), and by format of risk feedback (NM versus PT, x²=0.13, p=0.72) (Supplemental Table 1). Similarly, physician discussion did not vary significantly between all four randomization groups (x²=3.78, p=0.29), by type of feedback (GRS+FH versus FH, x²=3.41, p=0.065) or by format of feedback (NM versus PT, x²=0.37, p=0.54) (Supplemental Table 1).

Three years after results disclosure, medical records showed 33% (n=160 of 492) of participants had undergone PSA screening. Analyzing only subjects with an observation, this rate of PSA screening at three years did not significantly vary between all four randomization groups (x²=3.78, p=0.29), by type of feedback (GRS+FH versus FH, x²=1.7, p=0.19) or based on format of the risk feedback (NM versus PT, x²=0.61, p=0.44) (Supplemental Table 1).

Effects of Lifetime PCa Risk on Anxiety, PSA Discussions, & Screening

In the GRS+FH arm, as participants were given increased lifetime PCa risks, they were significantly more likely to report PSA discussion with a physician by three months (Wald x²=9.11, p=0.0025) (Figure 3a), engage in PSA screening by three months (Wald x²=6.13, p=0.013) (Figure 3b), and engage in PSA screening by three years (Wald x²=9.7, p=0.0018) (Figure 3c) (Supplemental table 2). For example, among men whose given PCa risk were <1, 1-2, 2-3, and >3-fold higher than population average, we observed 2.7%, 4.8%, 9.4%, and 16.7% of these men elected PSA screening by three months, respectively. Among subjects with a negative family history in the GRS+FH arm (n=229), given risk was significantly associated with PSA screening (n=77, 33.6%) at 3 years (Wald x²=8.9, p=0.0027). In contrast, none of these outcome trends for PSA screening and physician discussion were observed in the FH arm (Figure 3a-c and Supplemental table 2). Given risk was significantly associated with post-result anxiety levels indicative of clinical importance (STAI S-Score pro-rated cutoff of 41 or greater^24,25) in the GRS+FH arm (Wald x²=24.9, p<0.0001), but not in the FH arm (Wald x²=2.77, p=0.09).

Figure 3. Participant health behaviors, in FH and GRS+FH arms, stratified by given risk.

Black bars represent the percentage of participants that reported engaging in health behaviors during three month follow-up. Blue bars represent these same percentages, but stratified by category of given risk. Risk categories are <1, 1-2, 2-3, and >3-fold higher than the population average of 18.3% (categories were 0-17%, 18-35%, 37-54%, 55-80%, respectively)

a. Discussed PSA screening with physician per self-report during 3 month follow-up.

b. Engaged in PSA screening per self-report during 3 month follow-up.

c. Engaged in PSA screening, per medical record review, during 3 years of follow-up.

Post-Result Anxiety and Pre-Post Change in Anxiety by Group

Immediate post-result anxiety did not significantly differ by randomization group (x²= 2.39, p=0.49), by feedback type (GRS+FH versus FH, x²=0.93, p=0.34) or format (NM versus PT, x²=1.46, p=0.23) of risk feedback (Supplemental Table 3). From immediate pre-result (mean= 15.66) to immediate post-result (mean= 15.31), the average anxiety score for the complete study sample decreased significantly (S=−11660, p=0.0007) (Supplemental Table 3). The change in anxiety scores did not vary significantly across all four randomization groups (x²=2.18, p=0.54), by feedback type (GRS+FH versus FH, x²=0.93, p=0.33) or risk feedback format (NM versus PT, x²=0.081, p=0.78) (Supplemental Table 3). Although no group differences were observed in baseline anxiety level (Supplemental Table 3), we ran linear regression adjusting for baseline and still observed no group differences in pre-post anxiety.

When anxiety levels indicative of clinical concern were evaluated (STAI S-Score pro-rated cutoff of 41 or greater^24,25), group differences were observed in the number of subjects with clinically important anxiety level at baseline; linear regression adjusting for baseline anxiety showed no group differences in post result clinically important anxiety. Comparing the 104 subjects that exceeded this threshold post-results versus the remaining 590 subjects, there was no significant difference in rate of PSA discussion with a physician by three months (x²=0.018, p=0.89) or engaging in PSA screening by three months (x²=0.018, p=0.89) (Figure 3b); however, this subset of subjects was significantly more likely to engage in PSA screening by three years (x²=3.99, p=0.046). Regression analysis adjusting for baseline showed no significant relationship between post result clinically important anxiety and post-result PSA screening by three years.

Recall of Risk

As a measure of comprehension, we evaluated recall of the risk given to subjects (i.e. given risk). We observed a significant linear relationship between given risk and recalled risk at the immediate post-results assessment (β=93.5, s.e.−1.55, t=60.2, p<0.0001) (Supplemental Figure 2a), and at the 3-month follow-up assessment (β=94.7, s.e.=2.93, t=32.4, p<0.0001) (Supplemental Figure 2b). At both assessments, the relationship between given risk and recall was significant in each of the four randomization groups (p<0.001).

Discussion

The prospective, randomized study design allowed for comparison of counseling regarding individual lifetime risk of PCa based on GRS+FH vs. FH. The observed rates of PSA screening (3.14% at 3 months, 33% by 3 years) were close to the expected three-month rate (3.14%) and three-year rates (37.5%) based on published national rates for men age 40-49 during our study timeframe.²⁶ We found no significant differences in the overall rate of PSA screening between study groups after risk assessment, suggesting neither the source of risk feedback (genetic or family history) nor the presentation format (numerical only or numerical plus pictograph) were important determinants of screening behavior.

The level of given risk in the GRS+FH arm was a significant determinant of screening behavior. At 3 months, participants in the GRS+FH arm that were given >3-fold greater risk were 2.2 times more likely to discuss PSA screening with a physician, and 3.8 times more likely to engage in PSA screening compared with those given average risk, providing evidence of increased screening utilization among individuals that were given higher-risks. Similar trends were observed upon review of 3-year medical record data. The observation that given risk was significantly associated with PSA screening at 3 years among subjects with a negative family history in the GRS+FH arm further underscores the potential for GRS+FH to target PSA screening amongst a group of men typically considered at uniformly low risk. Our observations of genomic-targeted PSA screening are relevant to the PSA screening debate. A major concern is overtreatment of indolent PCa, in part because current screening and risk assessment tools do not adequately differentiate indolent versus clinically significant PCa at an early stage. Current validated SNPs and/or FH are believed predictive of any PCa rather than high-grade PCa; coupled with PSA screening, this may lead to increased diagnosis of early stage PCa that would never require treatment. However, there is growing evidence that risk-based targeting of PSA, including GRS, may reduce overdiagnosis. Overdiagnosis is the detection by screening, of prostate tumors that would not have presented clinically in a person’s lifetime⁵. A reduction in overdiagnosis is important because it means clinically relevant PCa should comprise a greater proportion of diagnoses, and this could plausibly reduce overtreatment and mortality. A recent retrospective analysis of 9,404 PCa cases and 7,608 controls in the United Kingdom found targeting PSA screening to men at higher multi-marker genetic risk would result in a 56% decrease in overdiagnosis.²⁷ Similarly, a retrospective analysis of 4967 men in the Finnish Prostate Cancer Screening Trial reported a potential 37% decrease in overdiagnosis if PSA screening had been offered only to men at above average risk based on a multi-marker panel of PCa risk variants.²⁸ By reducing overdiagnosis, GRS targeting of PSA screening therefore stands to provide additional opportunities for consideration of treatment options among men that have greater a-priori need for such treatment.

A prior multi-marker genetic risk assessment study reported no differences in diet, exercise, or use of screening tests, from baseline to 5.6 months average follow-up.²⁰ Another recent randomized trial of genetic and environmental risk assessment for colorectal cancer did not observe screening differences within six months by group or given risk level.²⁹ That study utilized subjects that were already non-compliant with screening, whereas we focused on screening-naïve subjects. The present study is likely the first prospective randomized trial to observe screening behavior changes, confirmed by medical records, in response to multi-marker genetic risk feedback in an average-risk population. Our findings of targeted screening are similar to prior studies of high-penetrance genes, with changes in mammography rate following disclosure of BRCA1 results.^30,31 and in colonoscopy associated with HNPCC genes.³² The finding of behavioral change associated with increasing risk in the GRS+FH arm may reflect a combination of a prospective randomized design, large study population, retention of 97% participants until 3 month follow-up, confirmed medical record observations on 70% by three years, focus on PCa, prediction of cancer risk, characteristics of the study population, and the broad spectrum of risk stratification.

Studies of hereditary PCa have shown that risk increases with additional affected family members^33-35. However, at study commencement, major clinical guidelines for PSA screening in the U.S. considered FH as either positive or negative. More importantly, prior large prospective studies in the general population, that are most useful for precise and accurate risk model estimates, have collected family history as a binary variable. Accordingly, FH risk estimates were binary in the present study, 17% negative versus 23% positive. This binary approach allowed for accurate and stable risk estimates to all subjects, while reflecting the binary FH risk assessment used to guide PSA screening in primary care settings where such screening often occurs. Unfortunately, binary FH risk assessment limits risk stratification, and as seen in our study, limits the potential to motivate behavior change. GRS+FH adds information from genetic markers to FH, distributing risk from 0% to 80% in this study. The importance of the larger range of risk feedback from GRS+FH is highlighted by our observation that the level of given risk was critical in explaining screening behaviors in the GRS+FH arm. In light of our findings, clinicians and future studies should consider incremental rather than binary assessment of family history risk for PCa, consistent with the latest PSA screening guidelines from the ACS.³ Importantly, only 9% of subjects in our study reported a positive family history. GRS+FH could reduce misclassification of risk for patients in which FH is negative (77% in the present study) or not available (14% in the present study).^36-39 FH and GRS+FH are complimentary, neither is diagnostic, and both can stratify risk to enable targeted screening.

Based on our results, a subset of patients will experience clinically important levels of anxiety in connection with the disclosure of risk feedback; clinicians and laboratories should therefore ensure individuals who undergo risk assessment and testing receive adequate pre- and post-results counseling. Overall, the provision of risk feedback led to statistically significant decreases in average anxiety, consistent with genetic testing studies of colon cancer and Alzheimer’s disease.^29,32,40 The provided risk feedback was accurately recalled, and anxiety modestly increased (although not clinically relevant) in direct proportion to given risk, similar to previous multi-marker genetic studies.^20,41 These findings add to the evidence suggesting concerns of providing multi-marker genetic feedback to patients may be over-estimated.

The lack of differences in recall associated with format of risk feedback conflicts with findings from several prior studies that indicated pictographs are a superior method to convey risk feedback to patients across all numeracy levels.²¹ However, much of the prior work in risk communication has utilized hypothetical scenarios in controlled laboratory studies rather than personal risk results delivered to patients in clinical settings.⁴² A recent empiric study on communication of breast cancer risks is consistent with our findings of no observable benefit in recall when using pictographs to communicate risk.⁴² Given that clinical genetic results are conveyed in a variety of formats, our findings may help inform clinical practices in risk communication, as well as future research. Perhaps there is no single approach for the communication of genetic risks that is beneficial across all numeracy levels and in all settings; in this case, it will be crucial to identify subsets of patients and clinical settings in which different methods have measurable benefit.

We made great efforts to provide consistency in all aspects of our study interactions with study subjects, including a genetic counseling script and handouts that directly addressed prostate cancer and screening to detect prostate cancer. The third sentence in the script said, “Because this is a research study, I will be sticking to a pre-written script to make sure I provide you with the same information as all of the other men in this study.” Certified master’s level genetic counselors were utilized to help ensure quality and consistency. In contrast, we intentionally made no effort to influence interactions that were external to the study, such as those with primary care providers (PCP). This allowed us to evaluate the occurrence of PCP appointments and uptake of PSA screening as outcomes in a way that more closely reflects the outcomes that might be observed in typical clinical settings. This approach may improve translation of our results to clinical settings in which PSA screening typically takes place, and in which variation in practice and individual decision-making is to be expected.

Our study approach did not assess variables at intermediate points between three months and three years. When designing the study, we considered additional surveys, deciding that fewer surveys would minimize the potential influence on participants’ longer term behaviors. Fortunately, the comparable rates of PSA screening in our study and the general population would suggest our measures did not have a significant impact on PSA screening behaviors at three months or three years.

The cost versus benefit of the GRS targeted approach utilized in this study is challenging to assess, but key factors can be described. Similar to other screening approaches, many genetic tests would be performed in order to identify the smaller subset of men at high risk; the added costs of the genetic test should be weighed against the reduced cost of PSA screening tests avoided. Commercial offerings of multi-marker multi-disease genetic tests have price points similar to commonly used screening tests and are conducted once in a lifetime, while PSA screening is typically repeated many times over a man’s life (up to annually), creating an additional cost differential. In addition, if GRS-targeted PSA screening reduces overdiagnosis and overtreatment as described above, significant savings should be derived from the large majority of men with indolent PCa that would not undergo PSA screening, nor would they incur costly biopsies or treatments. Future interventions could positively influence the cost-benefit ratio further by more actively encouraging or discouraging PSA screening based on strata of risk result (e.g. recommend PSA for the 12.5% (n=44/350) of subjects with >2 fold risk by GRS), rather than the passive patient-driven approach we utilized. Considering these factors, there would seem to be significant benefit to GRS in this setting, but formal cost-benefit analysis is needed prior to widespread use.

Limitations of this study include the inability to evaluate PCa detection rates and outcomes of PCa treatment based on low (expected) event rate during the follow-up period. Demographic features of the study population may limit generalization. Some subjects were lost to follow-up. Despite the limitations, these results indicate how genetic testing results are perceived and acted upon, for a method that could reduce overdiagnosis.

In summary, in this prospective RCT of 700 men, provision of individual GRS+FH did not lead to significant increases in anxiety or use of PSA screening. Rather, GRS+FH led to targeted use of PSA screening among higher risk men, which may improve PSA screening performance.