Author's Accepted Manuscript Intermediate-term Risk of Prostate Cancer is Directly Related to Baseline PSA: Implications for Reducing the Burden of PSA Screening Jonathan Gelfond , Kara Choate , Donna P. Ankerst , Javier Hernandez , Robin J. Leach , Ian M. Thompson , Jr.

PII: DOI: Reference:

S0022-5347(15)00292-X 10.1016/j.juro.2015.02.043 JURO 12220

To appear in: The Journal of Urology Accepted Date: 10 February 2015 Please cite this article as: Gelfond J, Choate K, Ankerst DP, Hernandez J, Leach RJ, Thompson IM Jr., Intermediate-term Risk of Prostate Cancer is Directly Related to Baseline PSA: Implications for Reducing the Burden of PSA Screening, The Journal of Urology® (2015), doi: 10.1016/ j.juro.2015.02.043. DISCLAIMER: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our subscribers we are providing this early version of the article. The paper will be copy edited and typeset, and proof will be reviewed before it is published in its final 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.

Embargo Policy All article content is under embargo until uncorrected proof of the article becomes available online. We will provide journalists and editors with full-text copies of the articles in question prior to the embargo date so that stories can be adequately researched and written. The standard embargo time is 12:01 AM ET on that date. Questions regarding embargo should be directed to [email protected].

ACCEPTED MANUSCRIPT

Intermediate-term Risk of Prostate Cancer is Directly Related to Baseline PSA:

Jonathan Gelfond, M.D., Ph.D.1,4 Kara Choate, B.S.

SC

Donna P. Ankerst, Ph.D.1,2,4

RI PT

Implications for Reducing the Burden of PSA Screening

Javier Hernandez, M.D. 2,4

M AN U

Robin J. Leach, Ph.D. 2,3,4

Ian M. Thompson, Jr., M.D.2,4

From the Departments of 1Biostatistics and Epidemiology, 2Urology, 3

Cellular and Structural Biology and the 4Cancer Therapy and Research Center,

Correspondence to:

TE D

University of Texas Health Science Center at San Antonio, San Antonio, TX

Ian M. Thompson, Jr., M.D.

EP

U627, Office of the Director

AC C

Cancer Therapy and Research Center 7979 Wurzbach Rd San Antonio, TX 78229 210-450-1408 Fax 210-450-1100 [email protected]

Supported by grants: U01 CA086402 (IMT, RJL, DPA, JH), P30 CA054174 (IMT, RJL, JG)

ACCEPTED MANUSCRIPT

Abstract Purpose: Prostate-specific antigen (PSA) screening is controversial, as a large number of men must be

RI PT

screened annually to achieve a benefit. We sought to determine if baseline PSA could reliably predict subsequent risk of prostate cancer (PCA) and risk of consequential PCA. Materials and Methods:

SC

A multi-ethnic cohort of 2923 PCA-free men, were recruited between 2000 and 2012 and

followed for a median of 7.5 years Baseline PSA was stratified into 6 strata and relative hazards

M AN U

of PCA detection for each PSA strata were estimated, adjusting for ethnicity, family history, and age. Results:

There were 289 cases of PCA diagnosed in patients during follow-up. Men with baseline PSA in the lowest stratum PSA [0.1 to 1.0 ng/mL] were at greatly reduced risk of PCA during follow-up.

TE D

this half of the cohort with PSA ≤1.0 ng/mL had a 3.4% (95% CI [2.1, 4.5]). 10-year risk of PCA; 90% of the cancers were low-risk. By comparison, the other half had a 15 to 39% risk of cancer detection with a 39% risk in the highest stratum (3-10 ng/mL).

EP

Conclusions:

Optimal PSA screening frequency for men with PSA levels of 0.1 – 1.0 ng/mL may be up to

AC C

every 10 years. This approach has the potential to dramatically reduce the cost of screening, reducing overdetection of inconsequential tumors, while maintaining detection of tumors for which treatment has been proven to reduce PCA mortality.

ACCEPTED MANUSCRIPT

Introduction

Screening for prostate cancer (PCA) with serum prostate-specific antigen (PSA) is a

RI PT

controversial strategy for reducing suffering from a disease that causes almost 30,000 deaths annually in the United States.1 Because of the ubiquity of low-grade tumors, commonly

detected during screening, even in the only clinical trial of screening to show a mortality

SC

reduction, the large number of patients needed to screen and treat to prevent one PCA death has led some to conclude that the burden of screening and treatment may be too great to justify

M AN U

this approach. 2

PSA testing is generally conducted annually. While the advantage of this approach is that it facilitates adherence to follow-up and will identify the occasional patient with prostate cancer whose PSA rapidly increases; such frequent testing leads to multiple ‘normal’ test values in

TE D

most men (an inefficient use of resources) as well as many ‘spikes’ in PSA (that often return to normal), prompting biopsies, that often identify low-risk, often-inconsequential tumors. 3

EP

As evidence is incontrovertible that small, low-grade tumors found at the time of autopsy in men who never had a clinical diagnosis nor symptoms from PCA can be found when prostate biopsy

AC C

is performed, the risk of ‘overdetection’ and subsequent overtreatment of these tumors is substantial. 4 Evidence of this is perhaps best found in the two large, randomized PSA screening studies. In the U.S., intensive PSA testing did not reduce the risk of PCA death when compared with ‘usual’ community practice. 5 In Europe, the European Randomized Study of Prostate Cancer (ERSPC) screening found mixed results of screening: with 13 years of followup, 781 men had to be invited for screening and 27 cancers had to be diagnosed to prevent one PCA death. 2 Widespread screening with PSA has improved diagnosis of PCA by facilitating early diagnosis, prior to metastasis. However, in so doing, as noted in the ERSPC trial,

ACCEPTED MANUSCRIPT

screening results in many unnecessary visits as well as detection of inconsequential tumors. In the U.S. Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) , the 4-year risk of a false positive PSA was 10.4%; the risk of a false positive digital rectal examination

RI PT

(DRE) was 15%. 6 As a result of risk of overdetection and overtreatment, the United States Preventive Task Force gave PSA screening for PCA a grade D recommendation (i.e.,

recommend against screening) in 2012. 7 Recognizing that annual screening may be too

frequent, the 2013 American Urological Association guidelines recommended men aged 55–69

SC

be offered biennial screening in the setting of shared decision-making. 8 The implications of a

M AN U

reduced frequency of screening could be imputed from the estimated cost of PSA screening and subsequent procedures in Medicare beneficiaries aged 66 to 99 years; in this population, annual expenditures were $145 million, approximately one-third of total Medicare spending on PCA screening. 9

TE D

The use of a ‘baseline’ PSA level has been demonstrated to be predictive of long-term PCA diagnosis and mortality. 10-12 In a 1756-subject Swedish PCA screening trial, among men with a PSA < 2.0 ng/mL, screening resulted in a significant increase in PCA incidence without an

EP

impact on subsequent PCA mortality. 13 We sought to determine the utility of a baseline PSA

AC C

level for predicting the frequency of PSA testing in a large, multi-ethnic U.S. cohort of men.

ACCEPTED MANUSCRIPT

Methods Subjects The San Antonio Biomarkers Of Risk of PCA study (SABOR) is a National Cancer Institute Early

RI PT

Detection Research Network sponsored Clinical Validation Center comprising more than 4,000 men from the San Antonio and South Texas area without a prior diagnosis of PCA. Cohort

enrollment began in 2000; between 2000 and 2012, subjects were seen annually for PSA and

SC

DRE testing with an average of approximately 6 visits per participant. Since 2011, subjects were followed with PSA only and those with a PSA < 1.0 ng/mL were seen biannually.

M AN U

Extensive baseline demographic, behavioral, and medical data were collected. Prostate biopsy was recommended based on community standards. Over the course of follow-up, over 340 cases of PCA have been diagnosed in the SABOR cohort. We restricted the analytical data to patients who reported race or ethnic categories as white non-Latino, black non-Latino or white

TE D

Latino, that had at least 1 month follow-up time, and who had initial PSA ≤ 10 ng/mL.

To evaluate the potential impact of PSA testing on risk of PCA progression and death, we established two definitions of low-risk PCA. The most limited definition was Gleason 3+3

EP

disease in 3 or fewer biopsy cores, none with more than 50% of one core. A broader definition included all of the subjects in the first but was expanded to include subjects with Gleason 3+4

AC C

disease in < 25% of up to two cores or up to 4 biopsy cores with Gleason 3+3 disease.

Statistical Analysis

Patient characteristics were contrasted among white non-Latinos, Latinos and black non-Latino participants using the chi-square test for count measures and the Mann-Whitney test for numerical measures.

ACCEPTED MANUSCRIPT

Prostate cancer-free survival was defined as the time between the first PSA measure and until PCA diagnosis. For patients who never developed PCA, censoring occurred at the latest recorded PSA level. Kaplan-Meier curves were used to estimate the cancer-free survival time,

RI PT

and strata were defined based on PSA intervals (0.0-1.0, 1.1-1.5, 1.6-2.0, 2.1-2.5, 2.6-3.0, 3.110.0 ng/mL). Kaplan-Meier plots were computed for each ethnic/racial group, and the Cox

proportional hazard model was fit to adjust for effects of age, family history (PCA in first degree

SC

relative), and ethnicity/race. Ten-fold cross validation was used to evaluate the discrimination ability (c-index) of multivariable Cox regression models predicting the time until PCA. The event

M AN U

of time-to-high-grade cancer (Gleason score > 6) detected on first diagnosis was considered as an alternative endpoint in a KM analysis. Low-grade cancers on first diagnosis were treated as censored in this analysis. This analysis targets aggressive cancers that are detected upfront in a screening population. All analyses were performed with the R environment for statistical computing (v3+, Vienna, Austria), and all statistical tests were performed at the two-sided 0.05

AC C

EP

TE D

level of statistical significance.

ACCEPTED MANUSCRIPT

Results

Table 1 displays demographic features of the 2923 SABOR participants who met the inclusion

RI PT

criteria. Of these, 289 (9.9%) were diagnosed with prostate cancer during follow-up. Notable is the broad racial/ethnic representation with white non-Latino, white Latino, and black non-Latino men well represented at 55%, 32%, and 13%, respectively. Consistent with the South Texas

SC

and congruent with the U.S. population, a substantial fraction of men were overweight: 75% of all racial/ethnic groups had a BMI > 25. While overall median PSA was 0.9 ng/mL, the lowest

M AN U

median PSA was noted in black non-Latino men. Median follow-up was 7.5 years (range: 1 month to 12.4 years). Over 20% of subjects had at least one first-degree family member with a history of PCA.

Figure 1A displays time until first diagnosis of PCA according to baseline PSA. Consistent with

TE D

other U.S. studies, more than half of the men in this cohort were in the lowest stratum with an initial PSA of 1.0 ng/mL or less. 14 Immediately apparent is the low rate of PCA detection among this group, which was 3.4% (95% confidence interval (CI) [2.1, 4.5]). By contrast, men in the

EP

upper strata of baseline PSA between 1.1 and 10.0 ng/mL had substantially higher PCA risk. Specifically, 10-year PCA risks for PSA ranges of 1.1-1.5 and 3.1-10.0 ng/mL were 14.9% (95%

AC C

CI [10.1, 18.9]) and 39.0% (95% CI [31.8, 45.4]), respectively (Table 2). A comparison of 5- and 10-year risks is provided in Table 2 of the Supplementary Appendix. The 5-year PCA risk for the lowest PSA stratum was only 1.9% (95% CI [1.1, 2.6]). Table 3 shows the effects of PSA on PCA risk adjusted for age, ethnicity, and family history. Adjusted hazard ratios of the upper 5 PSA strata range from 4.0 to 16.9. Figures 1B, C, D display similar curves separately for the three ethnic/racial groups.

ACCEPTED MANUSCRIPT

We examined the ability of the Cox model with covariates family history, PSA stratum, age, and race group to predict time-to-PCA diagnosis and obtained a cross-validated c-index of 0.81. Separate models for the white non-Latino, white Latino, and black non-Latino obtained c-indices

RI PT

of 0.81, 0.83, and 0.87, respectively, suggesting that baseline PSA is a good predictor of timeto-PCA diagnosis.

SC

The 5-year risks of PCA in the lowest PSA stratum for white non-Latinos, Latinos, and black non-Latinos were 1.2%, 3.1%, and 1.6%, respectively (Table 3, Supplementary Appendix).

M AN U

Using the Cox proportional hazard model with family history and age, an interaction between race/ethnicity and PSA strata on PCA risk was borderline significant (p = 0.04). These results are shown in Table 5 of the Supplementary Appendix. The effect of PSA stratum trends towards being less in Latinos compared to White non-Latinos. Also, the effect of the highest PSA stratum 3-10 ng/mL trends toward being higher (HR = 2.8, 95% CI [0.9, 8.3], p=0.07) in black

TE D

non-Latinos relative to white non-Latinos. Further, black subjects are more likely to have low PSA 0-1.0 ng/mL (62.4%) than the other groups (White 51.9% and Latino 58.2%), as well as less likely to be in the highest PSA interval 3-10 ng/mL (5.6% versus White 10.5% and Latino

EP

11.4%) (Table 4, Supplementary Appendix). Although the limited sample sizes preclude a definitive conclusion, with the exception of Black, non-Latino men (in whom the lowest strata

AC C

may have a higher risk), lower PSA levels predicted a low intermediate-term risk of subsequent PCA diagnosis.

We also examined the risk of high-grade PCA (Gleason Score >6) over time. The lowest PSA stratum had 5- and 10-year high-grade PCA risks of 0.46% (95% CI [0.09, 0.85]) and 0.91% (95% CI [0.33, 1.48]) (Table 6 Supplementary Appendix), respectively. While this indicates the observed risk under frequent screening (1-2 years) is quite low, these estimates may be negatively biased relative to a less frequent screening regimen.

ACCEPTED MANUSCRIPT

As the implication of these data relate to the potential of a baseline PSA to select men who could be considered for less-frequent PSA testing, we examined the characteristics of the 39

RI PT

cancers diagnosed among men with PSA ≤ 1.0 ng/mL at baseline. Of these subjects, 21 (54%) had a family history of PCA. The mean PSA levels at baseline and at diagnosis were 0.7 ng/mL and 2.4 ng/mL, respectively; 13 biopsies were prompted by an abnormal DRE and the

SC

remainder by a combination of changes in PSA and/or family history of PCA.

M AN U

Of the 39 men with PSA ≤ 1.0 ng/mL, 28 (72%) met the limited criterion of low-risk disease at the time of biopsy. Including men with the more expanded criterion, 35 (90%) could be considered having low-risk disease.

Eight of these 39 men remain on surveillance without progression. The remainder either initially

TE D

or ultimately received definitive treatment. Of the 39, two were diagnosed with Gleason grade tumors greater than 3+4: Gleason 4+4 in 7mm of one core in one subject and Gleason 5+4 and 4+5 in 80% of the total cores in another subject. The first of these subjects was treated with

EP

brachytherapy and lost to follow-up. The second subject whose PSA rapidly rose from 0.6 ng/mL to 9.6 ng/mL before diagnosis, underwent radical prostatectomy with a finding of Gleason

AC C

3+5+4 disease in 15% of the gland with seminal vesicle invasion and negative lymph nodes. The Gleason scores distributions of the other PSA strata are similar (Fisher’s test, p = 0.11) and provided in Supplemental Table 1. About 40% of all PSA strata have Gleason scores >6.

ACCEPTED MANUSCRIPT

Discussion

PSA testing has become a highly-controversial subject at a national level in the United States,

RI PT

primarily due to the discordant conclusions of two large screening trials, risk of overdetection of highly-prevalent low-grade tumors, side effects of definitive treatment, and the significant burden of either treatment or active surveillance. It is primarily due to these issues that the US

SC

Preventive Services Task Force recommended against PSA testing.

M AN U

An opportunity to reduce the burden of PSA testing could be to measure initial PSA and then develop a risk-adjusted frequency for testing. While risk of PCA is affected by other factors (family history, age, etc.), our data provide a strong rationale for a significant reduction in frequency of screening in men whose initial PSA is 1.0 ng/mL or less. Among these men, few PCAs are detected over a period of up to 12 years and, of cancers detected, the majority are

TE D

low-risk. It could be argued that the overall impact of detecting low grade tumors is a potential net-harm to the patient as the benefit of detection is small while the harm (including need for repeated monitoring and biopsy if surveillance is selected, significant negative impact on quality

EP

of life if definitive treatment is selected, and a prolonged period of cancer-induced anxiety) is

AC C

measurable and consequential.

It could be argued that with a risk-adjusted PSA testing strategy in which major reductions in frequency of PSA testing (e.g., every 10 years) is pursued, some PCAs of consequence would be missed; our data show this could occur. However, it is clear that even with annual or biannual PSA testing in the two prospective screening studies, some PCAs are missed; even in the case of the European screening study, the absolute reduction in PCA mortality risk was small.

ACCEPTED MANUSCRIPT

Our data provide a compelling rationale for less-frequent PSA testing in men whose PSA is ≤1.0 ng/mL. An option in these men may be to harmonize their PSA testing with other preventive medicine testing such as colonoscopy. While moving from an annual to an every-decade

RI PT

testing strategy in these men would be logistically more challenging (it is easier to check a PSA lab slip annually than remembering to do so once a decade), the widespread adoption of

electronic medical records and reminder ‘prompts’ to physicians should facilitate this process.

SC

These estimates of the PCA detection rates are limited to patients with similar demographic characteristics (age, race/ethnicity, etc.). The median study follow-up time of 7.5 years limits the

M AN U

certainty about long-term predictions. Further, it has been shown that predictive models that discretize continuous biomarkers into intervals can under- and overestimate patient-specific risks15 (e.g., PSA of 0.1 vs 1.0 ng/mL within the interval 0.0-1.0 could confer different risks). Future work should develop and validate patient-specific calculators that estimate the 5- and 10year risk probabilities based upon the patients’ risk factors (PSA level, ethnicity, family history,

TE D

age, etc.) to better guide adaptive screening plans. Also, the screening strategy should be a shared decision that includes the patient’s personal concerns/preferences. It is clear, however, that not adopting a risk-adjusted early detection program runs the risk of abusing limited health

risk tumors.

EP

care resources and causing unneeded treatments and adverse consequences to men with low-

AC C

Acknowledgements

The SABOR project is supported by the San Antonio Center of Biomarkers of Risk for Prostate Cancer U01 CA86402 and the men who participated in SABOR for their contributions to this program. We thank the reviewers whose thoughtful suggestions have improved this manuscript. This work is supported by grants: U01 CA086402 (IMT, RJL, DPA, JH), P30 CA054174 (IMT, RJL, JG)

ACCEPTED MANUSCRIPT

Figure I. Legend Kaplan-Meier curves of time-to-prostate cancer detection by PSA strata (by color) and race/ethnicity (by panel). Panel A includes all races and ethnicities. The other panels represent Black non-Latino (B), White non-Latino (C), and White Latino (D), respectively. The number at

RI PT

risk for each PSA stratum is given. The risk of prostate cancer detection clearly increases with

SC

PSA level, and long-term (10-year) risk of the lowest strata being small (less than 5%).

4. 5. 6. 7. 8. 9. 10.

11. 12.

TE D

3.

EP

2.

Siegel, R., Ma, J., Zou, Z. et al.: Cancer statistics, 2014. CA: a cancer journal for clinicians, 64: 9, 2014 Schröder, F. H., Hugosson, J., Roobol, M. J. et al.: Screening and prostate cancer mortality: results of the European Randomised Study of Screening for Prostate Cancer (ERSPC) at 13 years of follow-up. The Lancet, 384: 2027, 2014 Ankerst, D. P., Miyamoto, R., Nair, P. V. et al.: Yearly prostate specific antigen and digital rectal examination fluctuations in a screened population. The Journal of urology, 181: 2071, 2009 Thompson, I. M., Ankerst, D. P., Chi, C. et al.: Operating characteristics of prostate-specific antigen in men with an initial PSA level of 3.0 ng/ml or lower. Jama, 294: 66, 2005 Andriole, G. L., Crawford, E. D., Grubb III, R. L. et al.: Mortality results from a randomized prostate-cancer screening trial. New England Journal of Medicine, 360: 1310, 2009 Croswell, J. M., Kramer, B. S., Kreimer, A. R. et al.: Cumulative incidence of false-positive results in repeated, multimodal cancer screening. The Annals of Family Medicine, 7: 212, 2009 Moyer, V. A.: Screening for prostate cancer: US Preventive Services Task Force recommendation statement. Annals of internal medicine, 157: 120, 2012 Carter, H. B., Albertsen, P. C., Barry, M. J. et al.: Early detection of prostate cancer: AUA Guideline. The Journal of urology, 190: 419, 2013 Ma, X., Wang, R., Long, J. B. et al.: The cost implications of prostate cancer screening in the Medicare population. Cancer, 120: 96, 2014 Lilja, H., Ulmert, D., Björk, T. et al.: Long-term prediction of prostate cancer up to 25 years before diagnosis of prostate cancer using prostate kallikreins measured at age 44 to 50 years. Journal of clinical oncology, 25: 431, 2007 Loeb, S., Carter, H. B., Catalona, W. J. et al.: Baseline prostate-specific antigen testing at a young age. European urology, 61: 1, 2012 Vickers, A. J., Cronin, A. M., Björk, T. et al.: Prostate specific antigen concentration at age 60 and death or metastasis from prostate cancer: case-control study. Bmj, 341, 2010

AC C

1.

M AN U

References

ACCEPTED MANUSCRIPT

RI PT

SC M AN U TE D

15.

EP

14.

Carlsson, S., Assel, M., Sjoberg, D. et al.: Influence of blood prostate specific antigen levels at age 60 on benefits and harms of prostate cancer screening: population based cohort study. BMJ: British Medical Journal, 348, 2014 Porter, M. P., Stanford, J. L., Lange, P. H.: The Distribution of Serum Prostate-Specific Antigen Levels Among American Men: Implications for Prostate Cancer Prevalence and Screening. The Prostate, 66: 1044, 2006 Bennette, C., Vickers, A.: Against quantiles: categorization of continuous variables in epidemiologic research, and its discontents. BMC medical research methodology, 12: 21, 2012

AC C

13.

ACCEPTED MANUSCRIPT

Table 1. Entry Demographics by Race and Ethnicity. White non-Latino

Latino

Black non-Latino

Total

Age N Mean (SD) Median [Q1, Q3] Min, Max

1606 60.3 (9.5) 59.9 [53.8, 66.9] 29, 88.5

945 56 (9.4) 55.6 [49.5, 62.4] 28.6, 85

372 52.3 (10) 50.8 [45.6, 59.2] 28.3, 81.8

2923 57.9 (10) 57.5 [50.8, 64.7] 28.3, 88.5

BMI N Mean (SD) Median [Q1, Q3] Min, Max

1026 28.3 (4.6) 27.6 [25.1, 30.5] 15.7, 52.6

768 29.6 (5) 28.8 [26.3, 32] 19.3, 54.5

269 29.7 (5) 29.2 [26.6, 32.6] 18.7, 48.9

2063 28.9 (5) 28.2 [25.7, 31.4] 15.7, 54.5

Start PSA N Mean (SD) Median [Q1, Q3] Min, Max

1606 1.5 (1.4) 1 [0.6, 1.8] 0.1, 8.9

945 1.4 (1.4) 0.9 [0.5, 1.6] 0.1, 9.4

372 1.2 (1.1) 0.8 [0.5, 1.4] 0.1, 8.9

2923 1.4 (1) 0.9 [0.6, 1.7] 0.1, 9.4

Followup time N Mean (SD) Median [Q1, Q3] Min, Max

1606 7.3 (3.6) 8 [4.1, 10.6] 0.1, 12.4

945 6.1 (3.6) 6.3 [2.1, 9.1] 0.1, 12.3

372 6.6 (3.6) 7 [3.1, 10] 0.1, 12.4

2923 6.8 (4) 7.5 [3.3, 10.1] 0.1, 12.4

Fam History No Yes Total

1210 (75.34) 396 (24.66) 1606

776 (82.12) 169 (17.88) 945

296 (79.57) 76 (20.43) 372

2282 (78.07) 641 (21.93) 2923

Abnormal DRE No Yes Total

1307 (81.38) 299 (18.62) 1606

819 (86.67) 126 (13.33) 945

336 (90.32) 36 (9.68) 372

2462 (84.23) 461 (15.77) 2923

SC

M AN U

TE D

EP

AC C

RI PT

Characteristic

p-value < 0.001

< 0.001

< 0.001

< 0.001

< 0.001

< 0.001

ACCEPTED MANUSCRIPT

Latino

Black non-Latino

Total

1427 (88.85) 179 (11.15) 1606

876 (92.7) 69 (7.3) 945

331 (88.98) 41 (11.02) 372

2634 (90.11) 289 (9.89) 2923

p-value

TE D

M AN U

SC

RI PT

0.01

EP

Case Negative Positive Total

White non-Latino

AC C

Characteristic

- Page 2 of 10

ACCEPTED MANUSCRIPT

Table 2. Ten-year risk estimate of prostate cancer detection from the Kaplan-Meier curve according to baseline PSA.

Ten-year risk (%)

95% confidence interval

1615 489 277 145 100 297

40 51 48 26 29 95

3.37 14.94 23.50 25.13 33.66 38.96

[2.24, 4.47] [10.83, 18.87] [16.88, 29.59] [15.81, 33.41] [21.58, 43.88] [31.76, 45.40]

TE D

M AN U

SC

RI PT

N events

EP

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

N

AC C

PSA range (ng/mL)

- Page 3 of 10

ACCEPTED MANUSCRIPT

0.01 0.12 0.21 0.22 0.26 0.25 0.20 0.18 0.14

HR

95% confidence interval

p-value

1.02 1.94 4.04 6.89 8.09 12.44 16.91 1.75 0.84

[1, 1.03] [1.52, 2.47] [2.67, 6.13] [4.51, 10.54] [4.89, 13.4] [7.64, 20.26] [11.45, 24.98] [1.24, 2.49] [0.63, 1.11]

0.02 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 0.002 0.21

SC

0.016 0.663 1.397 1.930 2.091 2.521 2.828 0.562 -0.180

TE D

M AN U

Standard Error

EP

Age at baseline Family History (Yes vs. No) PSA stratum(1,1.5] PSA stratum(1.5,2] PSA stratum(2,2.5] PSA stratum(2.5,3] PSA stratum(3,10] Race Black non-Latino (vs White) Race Latino (vs White)

Log(HR)

AC C

Characteristic

RI PT

Table 3. Cox proportional hazard regression estimates for prostate cancer detection fitted to 2923 SABOR participants (289 cases of cancer detection). Reference levels: Family History = No, PSA stratum = [0.0, 1.0], Race Group = White.; HR: hazard ratio.

- Page 4 of 10

ACCEPTED MANUSCRIPT

Supplemental Table 1. Table of Gleason Score distribution of cases by baseline PSA. G2

G4

G5

G6

G7

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10] Total

0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (3.6%) 0 (0%) 1 (0.4%)

1 (2.4%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (0.4%)

0 (0%) 0 (0%) 0 (0%) 1 (4%) 1 (3.6%) 0 (0%) 2 (0.7%)

32 (76.2%) 30 (62.5%) 29 (60.4%) 14 (56%) 13 (46.4%) 54 (60%) 172 (61.2%)

7 (16.7%) 13 (27.1%) 14 (29.2%) 6 (24%) 10 (35.7%) 32 (35.6%) 82 (29.2%)

G8

G9

Total

1 (2.4%) 4 (8.3%) 3 (6.2%) 3 (12%) 2 (7.1%) 1 (1.1%) 14 (5%)

1 (2.4%) 1 (2.1%) 2 (4.2%) 1 (4%) 1 (3.6%) 3 (3.3%) 9 (3.2%)

42 48 48 25 28 90 281

AC C

EP

TE D

M AN U

SC

RI PT

PSA (ng/mL)

- Page 5 of 10

ACCEPTED MANUSCRIPT

Supplemental Table 2. Comparison of 5- to 10-year risk estimates of prostate cancer detection from Kaplan-Meier curves according to baseline PSA. 10-year risk (%)

10-year risk 95% CI

3.37 14.94 23.50 25.13 33.66 38.96

[2.24, 4.47] [10.83, 18.87] [16.88, 29.59] [15.81, 33.41] [21.58, 43.88] [31.76, 45.4]

RI PT

[1.11, 2.58] [3.63, 8.29] [6.17, 13.89] [8.27, 21.16] [13.71, 31.05] [24.28, 35.46]

SC

5-year risk 95% CI

1.85 5.99 10.11 14.96 22.87 30.09

TE D

M AN U

5-year risk (%)

EP

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

AC C

PSA (ng/mL)

- Page 6 of 10

ACCEPTED MANUSCRIPT

Supplemental Table 3. Comparison of 5-year risk estimates of prostate cancer detection from Kaplan-Meier curves by baseline PSA and race. N events

833 284 165 95 61 168

20 35 33 21 16 54

Latino

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

550 145 78 35 29 108

14 7 11 1 9 27

Black non-Latino

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

232 60 34 15 10 21

M AN U

TE D

Risk (95% CI)

1.26 7.43 10.87 16.36 24.46 26.99

[0.44, 2.07] [4.06, 10.69] [5.68, 15.79] [7.74, 24.18] [12.4, 34.86] [19.69, 33.62]

3.08 1.78 7.54 3.12 20.26 26.48

[1.34, 4.79] [0, 4.20] [0.89, 13.74] [0, 8.97] [2.4, 34.85] [16.45, 35.31]

1.56 7.95 11.36 36.00 20.00 76.26

[0, 3.31] [0.15, 15.14] [0, 22.74] [0, 59.34] [0, 41.32] [41.98, 90.28]

EP

6 9 4 4 4 14

Risk (%)

RI PT

N

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

SC

PSA stratum

White non-Latino

AC C

Race Group

- Page 7 of 10

ACCEPTED MANUSCRIPT

Supplemental Table 4. Cross-tabulation of PSA strata and race. While significantly associated the differences are not large (p < 0.0001 ). Black

Total

232 (62.4%) 60 (16.1%) 34 (9.1%) 15 (4%) 10 (2.7%) 21 (5.6%) 372 (100%)

1615 (55.3%) 489 (16.7%) 277 (9.5%) 145 (5%) 100 (3.4%) 297 (10.2%) 2923 (100%)

RI PT SC

Latino 550 (58.2%) 145 (15.3%) 78 (8.3%) 35 (3.7%) 29 (3.1%) 108 (11.4%) 945 (100%)

TE D

M AN U

White 833 (51.9%) 284 (17.7%) 165 (10.3%) 95 (5.9%) 61 (3.8%) 168 (10.5%) 1606 (100%)

EP

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10] Total

AC C

PSA (ng/mL)

- Page 8 of 10

ACCEPTED MANUSCRIPT

Log(HR)

Standard Error

HR

95% confidence interval

p-value

PSA stratum (1,1.5] (Black non-Latino vs White) PSA stratum (1,1.5] (Latino vs White)

0.063 -1.035

0.60 0.54

1.06 0.36

[0.33, 3.43] [0.12, 1.03]

0.91 0.06

PSA stratum (1.5,2] (Black non-Latino vs White) PSA stratum (1.5,2] (Latino vs White)

-0.520 -0.550

0.71 0.49

0.59 0.58

[0.15, 2.37] [0.22, 1.52]

0.46 0.27

PSA stratum (2,2.5] (Black non-Latino vs White) PSA stratum (2,2.5} (Latino vs White)

0.389 -2.227

SC

RI PT

Supplemental Table 5. The interaction between PSA and race group. The effects of PSA stratum relative to white non-Latinos are given adjusting for family history and age with a Cox Model. Modeled N =2923. Number of events =289.

0.72 1.08

1.48 0.11

[0.36, 6.02] [0.01, 0.9]

0.59 0.04

PSA stratum (2.5,3] (Black non-Latino vs White) PSA stratum (2.5,3] (Latino vs White)

0.368 -0.270

0.73 0.55

1.44 0.76

[0.35, 6.02] [0.26, 2.22]

0.61 0.62

PSA stratum (3,10] (Black non-Latino vs White) PSA stratum (3,10] (Latino vs White)

1.023 -0.486

0.56 0.42

2.78 0.62

[0.93, 8.27] [0.27, 1.4]

0.07 0.25

AC C

EP

TE D

M AN U

Characteristic

- Page 9 of 10

ACCEPTED MANUSCRIPT

Supplemental Table 6. Comparison of 5 to 10-year risk estimate of aggressive disease (Gleason score > 6) from Kaplan-Meier curves according to baseline PSA. strata. 10-year risk (%)

10-year risk 95% CI

0.91 5.80 10.48 11.35 16.68 17.54

[0.33, 1.48] [2.95, 8.57] [5.32, 15.36] [4.08, 18.06] [6.07, 26.09] [11.2, 23.42]

RI PT

[0.09, 0.84] [0.81, 3.85] [1.12, 6.11] [1.05, 9.43] [3.87, 17.48] [7.47, 15.6]

SC

5-year risk 95% CI

0.46 2.34 3.65 5.34 10.93 11.63

TE D

M AN U

5-year risk (%)

EP

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

AC C

PSA (ng/mL)

- Page 10 of 10

ACCEPTED MANUSCRIPT

12

0.0

0.2

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

833 284 165 95 61 168

680 220 126 60 43 98

2

4

470 150 77 37 21 57

6 Followup year

8

22 8 2 1 0 2

10

12

RI PT

0.6 0.4

SC

0.2 0.0

0

2

4

111 28 11 3 4 1

6

8

5 2 0 0 0 0

10

12

Followup year

D

1.0

TE D

0.4

0.6

AC C

0.8

EP

1.0

C

0.8

1.0 10

175 44 21 5 7 5

0.8

8

232 60 34 15 10 21

0.6

6

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

0.4

4

33 11 2 1 0 6

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

0.2

2

811 245 120 51 38 87

0.0

1225 360 197 88 67 158

Followup year

0

B

M AN U

1615 489 277 145 100 297

No PCA Proportion

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

0

No PCA Proportion

No PCA Proportion

0.8 N at Risk

0.2

0.4

0.6

[0,1] (1,1.5] (1.5,2] (2,2.5] (2.5,3] (3,10]

0.0

No PCA Proportion

1.0

A

0

550 145 78 35 29 108

370 96 50 23 17 55

2

4

230 67 32 11 13 29

6 Followup year

8

6 1 0 0 0 4

10

12

ACCEPTED MANUSCRIPT

Definitions of key abbreviations: PCA is Prostate Cancer

AC C

EP

TE D

M AN U

SC

RI PT

PSA is Prostate Specific Antigen