1 Screening for Prostate Cancer: Understanding the Evidence, Differences in Screening Guidelines, and Future Directions American Council of Life Insurers Medical Section Annual Meeting February 24, 2013 Robert A. Smith, PhD American Cancer Society Atlanta, GA
2 Prostate Cancer Then, Now, and Looking Forward Historical trends in testing and disease burden What do the randomized trials tell us about the role of screening in reducing deaths? How do we reconcile differences in current guidelines between the ACS and the USPSTF? Is there a way forward?
3 The modern era of PSA testing for early prostate cancer detection occurred after 1987, following FDA approval for the use of PSA to monitor recurrence of prostate cancer
4 The uptake of PSA testing was significant after Stamey, et al. s publication in the NEJM Data on PSA testing rates derived from Medicare data showed an increase from 1.2% in 1988 to nearly 40% in 1994 Incidence begins to rise in 1988, and peaks in 1992 The percent of white men with a 1 st PSA peaked at 19% in 1992, and subsequently declined, which probably contributed to the decline in incidence afterwards
5 U.S. Trends in Prostate Cancer Incidence & Mortality, Incidence begins to rise in 1988, and peaks in 1992 Mortality also peaks in 1992
6 Widespread screening began in the absence of guidelines from Expert Groups ACS first issued a recommendation for screening in 1992, 4 years after the beginning of a rapid increase in incidence, and after age-adjusted incidence peaked. There was not consensus on the value of PSA screening for prostate cancer No experimental data--the NCI and a European Consortium initiated prospective randomized controlled trials in the early 1990s Observational data became the basis for arguing the pros and cons of prostate cancer screening until 2009
7 What did early prostate cancer incidence and staging data suggest an impact from screening? After 1987, there was a rapid increase in incidence After 1991 there was a decrease in distant stage disease ( not shown) A decrease in mortality by age indicating a calendar year effect
8 Other Evidence During the period between , nonexperimental studies were the basis for concluding that PSA testing: (1) reduced prostate cancer mortality; and (2) did not reduce prostate cancer mortality Geographical comparisons Case series Case control studies Modeling
9 Evidence from Geographic Comparisons aka Natural Experiments Compare trends in disease-specific mortality across countries or areas with different levels of PSA screening
10 Tyrol County versus Rest of Austria In 1993, PSA testing was made freely available to men aged 45 to 75 years in the Federal State of Tyrol, Austria. At least two thirds of all men in this age range have been tested at least once during the first 5 years of the study
11 US and UK Prostate Cancer Mortality Rates by Age, UK US Mortality peaked in the US & UK in 1990, but subsequently declined more rapidly in the US Lancet Oncol 2008; 9:
12 Screening and Treatment In US and UK SCREENING? TREATMENT? A. UK & US Age Adjusted Prostate Cancer Mortality Rates UK C. UK & US Prostate Cancer Treatment Trends US B. Ratio of US & UK Age Adjusted Incidence Rates A & B suggest a screening effect, but C demonstrates considerable differences in application of advances in therapy
13 Geographic Studies To be interpretable a geographic study must document not only screening trends but also treatment trends before and during the PSA era Example: Screening and adjuvant HT use in Seattle and Connecticut Screening frequencies were not that different in SEA and CT after the very early years Treatment patterns in SEA and CT were different Example: Screening and treatment in USA and UK More screening and more treatment in US than in UK
14 Results from the PLCO and ERSPC Prostate Cancer Screening Trials were Published in March, 2009
15 PLCO Trial Enrollment: Number of Centers: 10 Study Size: 76,693 Randomization Men ages years 38,343 men invited to annual screening with DRE (4 years) and PSA (6 years) 38,350 men received usual care Adherence: ~ 85% for DRE & PSA Contamination: PSA %; DRE % Criteria for positive test results PSA > 4.0 ng/ml DRE, nodularity, induration, etc. Minimum follow-up (from randomization) = 7 years Endpoint: Prostate specific mortality
16 PLCO Trial At 7 years of follow-up Screening: 2,820 cases Controls: 2,322 cases Rate ratio = 1.22 At 10 years of follow-up Screening: 3,452 cases Controls: 2,974 cases Rate ratio = 1.17 Conclusion-significant overdiagnosis
17 PLCO Trial At 7 years of follow-up Screening: 50 deaths Controls: 44 deaths Rate ratio = % CI, At 10 years of follow-up Screening: 92 deaths Controls: 82 deaths Rate ratio = % CI, Conclusion no evidence of benefit
18 PLCO Trial Some Limitations At 10 years, follow-up is only complete for 67% of study subjects At 10 years, median follow-up time is longer for the screening group (6.3 years) vs. controls (5.2 years) this is expected, but highlights that more follow-up time is needed Quantitative conclusions about overdiagnosis are tentative due to uneven follow-up between invited and control group, limited follow-up overall in the study At enrollment, > 40% had undergone screening within the previous 3 years Contamination in the control group (~ 50%) although anticipated, comparative analysis of annual vs. opportunistic screening is needed Contamination may be underestimated (estimated by annual survey) Low rate of biopsy of positive cases
19 ERSPC Trial Enrollment: Initiated in early 1990 s Participating Countries: Netherlands, Belgium, Sweden, Finland, Italy, Spain, Switzerland Study Size: 162,387 Randomization Men ages years 72,952 men invited to annual screening with PSA every 4 years, average 2.1 tests per subject that underwent screening 89,435 men received usual care Adherence: ~ 82% for one time screening for PSA Contamination: very low Criteria for positive test results Typically, PSA > 3.0 ng/ml Follow-up (from randomization) = ~ 9 years Endpoint: Prostate specific mortality
20 Cumulative risk from prostate cancer in the ERSPCS As of 12/2006 Average follow-up of 8.8 years 214 deaths in the screening group vs. 326 in the control group The adjusted rate ratio for prostate specific mortality was 0.80, (95% CI, , P = 0.04) Adjusted for non-compliance, RR = 0.73, (95% CI, )
21 ERSPC Trial Number Needed to Invite to prevent 1 prostate cancer death = 1410 (Mean 1.7 screening visits over 9 year period) Number Needed to Screen = 1068 Number Needed to Diagnose to prevent 1 prostate cancer death = 48
22 ERSPCS Trial Some Limitations Longer follow-up is needed The screening interval may have been too wide Some lingering questions about the influence of selection bias in countries where control group men were unaware they were in a study Were men in the control group with prostate cancer prerandomization identified through the registries? Were men in the invited group with prostate cancer prerandomization not allowed to opt out of the study? Some lingering questions about differential treatment in the invited vs. control group
23 Recent results from the Göteborg arm of the ERSPC 20,000 men born between Randomized 1:1, 10,00 in each arm PSA screening every 2 years, up to ~ age 70 vs. usual care Endpoint: Prostate cancer specific mortality
24 Cumulative Incidence of Prostate Cancer Mortality in the Screening Group vs. the Control Group The absolute cumulative risk reduction of death from prostate cancer at 14 years was 0 40% (95% CI ), from 0 90% in the control group to 0 50% in the screening group. The rate ratio for death from prostate cancer was 0 56 (95% CI ; p=0 002) in the screening compared with the control group. The rate ratio of death from prostate cancer for attendees compared with the control group was 0 44 (95% CI ; p=0 0002). Overall, 293 (95% CI ) men needed to be invited for screening and 12 to be diagnosed to prevent one prostate cancer death. Hugosson J, et al. Lancet Oncol 2010;11:
25 Prostate Cancer Screening: ACS (2010); USPSTF Recommendations (2012)
26 Prostate Cancer Screening in Average and High Risk Men: ACS (2010); USPSTF (2012) Recommendation ACS USPSTF PSA Testing The ACS recommends that asymptomatic men who have at least a 10-year life expectancy should have an opportunity to make an informed decision with their health care provider about whether to be screened for prostate cancer, after receiving information about the uncertainties, risks, and potential benefits associated with prostate cancer screening. The U.S. Preventive Services Task Force (USPSTF) recommends against prostate-specific antigen (PSA)-based screening for prostate cancer. This is a grade D recommendation.
27 Critique of the USPSTF s D Rating for Prostate Cancer Screening A D rating states that there is moderate or high certainty that this service has no net benefit or that the harms outweigh the benefits. It is fair to say that the balance of benefits and harms is open to reasonable doubt However, there have been a number of common critiques of the USPSTF report
28 Critique of the USPSTF s D Rating for Prostate Cancer Screening Definitive conclusions based on incomplete data The ERSPC report was an interim analysis, published before the pre-specified main follow-up time (11 years). The update (2 additional years of follow-up) reported that PSA testing significantly reduces prostate cancer mortality (RR = 0.79, 95% CI, ; P =.001) For an excellent critique, see Carlsson, S. Journal of Clinical Oncology, Vol. 30, No 21 (July 20), 2012: pp
29 Critique of the USPSTF s D Rating for Prostate Cancer Screening All-cause mortality vs. cancer specific mortality One of the USPSTF s key questions was whether PSA testing reduced all-cause mortality. The question of all-cause mortality reveals a fundamental misunderstanding about the design of a screening trial, which focuses on diseasespecific mortality, and the statistical power to measure all-cause mortality
30 Critique of the USPSTF s D Rating for Prostate Cancer Screening Combining data from incompatible trials It has become de rigor to conduct meta-analyses when data from multiple trials are available However, trials must be relatively comparable for meta-analysis to be informative Several of the trials (Quebec and Norrkoping) have serious methodological weaknesses, and the PLCO and ERSPC evaluated PSA testing under very different circumstances
31 Critique of the USPSTF s D Rating for Prostate Cancer Screening Drawing biologically implausible conclusions The USPSTF speculated that the 1 year screening interval in the PLCO may have been less effective than the 4 year screening interval in the ERSPC. This is nonsense. A less frequent screening interval can reduce the rate of harms, but it will not improve cancer outcomes.
32 Critique of the USPSTF s D Rating for Prostate Cancer Screening Erroneous conclusions when addressing timeto-event data The USPSTF stated, 48 men received treatment for every prostate cancer-specific death prevented. This is incorrect, since the number represents men diagnosed, not treated (many were on surveillance protocols). Further, number needed to treat is highly dependent on the duration of follow-up ERSPC at 11 years (NND = 37) Göteborg at 14 years (NND = 12)
33 Critique of the USPSTF s D Rating for Prostate Cancer Screening Overestimation of harms The USPSTF estimated the 30-day perioperative mortality rate after radical prostatectomy as 0.5%, a outcome based on Medicare data from (i.e., 20 years ago). Today, men 65+, who are at higher risk for adverse outcomes, constitute are a minority of the radical prostatectomy series Contemporary estimates based on all men are closer to 0.1%
34 PSA testing today While there is a sound basis to disagree with the USPSTF recommendations, PSA testing in the U.S. does not commonly adhere to any recommendation: PSA testing begins too early, stops too late, and too many men with limited life expectancy or significant co-morbidity are being tested. Little evidence that shared decision making is taking place Patients report that their doctors tend to encourage PSA testing Men with elevated PSA commonly are encouraged to undergo radical treatment
35 Treatment Limitations of Treatment for Prostate Cancer Concerns about slow growing disease, overdiagnosis, and the side effects of treatment have led to prospective trials comparing radical treatment to observation or active surveillance Veterans Admin Cooperative Urological Group Study (1995) Randomized Scandinavian Prostate Cancer Group Study No. 4 (2011) [Symptomatic cases] Prostate Cancer Intervention versus Observation Trial (PIVOT) (2012) [Screen detected cases] Prostate Testing for Cancer and Treatment (ProtecT) (2002)
36 Radical Prostatectomy vs. Watchful Waiting in Early Prostate Cancer (SPCG-4) N Engl J Med 2011;364: Between 10/89 2/99, 695 men with symptomatic, early prostate cancer were randomized to watchful waiting or radical prostatectomy. Follow-up through 12/2009 (15 years, median 12.8 years) Primary outcome was prostate cancer death 55/347 men in the surgery group died vs. 81/348 in the watchful waiting group, RR = 0.62, (95% CI 0.44 to 0.87; P = 0.01) Absolute difference = 6.1%
37 Cumulative Incidence of Death from Any Cause and Death from Prostate Cancer, SPCG-4) N Engl J Med 2011;364:
38 Radical Prostatectomy vs. Observation for Localized Prostate Cancer (PIVOT) N Engl J Med 2012;367: November, January men with localized, screen detected prostate cancer (mean age, 67 years; median PSA value, 7.8 ng per milliliter) assigned either to radical prostatectomy or observation Follow-up through January 2010 (median 10 years). Primary outcome: All-cause mortality Secondary outcome: Prostate cancer mortality
39 Pivot Trial: Kaplan-Meier Plots, (A) All Cause Mortality, and (B) Prostate Specific Mortality. Radical Prostatectomy Observation 171 of 364 men (47.0%) assigned to radical prostatectomy died vs. 183 of 367 (49.9%) assigned to observation (hazard ratio, 0.88; 95% confidence interval [CI], 0.71 to 1.08; P = 0.22; absolute risk reduction, 2.9 percentage points). Among men assigned to radical prostatectomy, 21 (5.8%) died from prostate cancer or treatment, as compared with 31 men (8.4%) assigned to observation (hazard ratio, 0.63; 95% CI, 0.36 to 1.09; P = 0.09; absolute risk reduction, 2.6 percentage points). N Engl J Med 2012;367:
40 Radical Prostatectomy vs. Observation for Localized Prostate Cancer (PIVOT) Radical prostatectomy was associated with reduced all-cause mortality among men with a PSA value greater than 10 ng per milliliter (P = 0.04 for interaction) and possibly among those with intermediate-risk or high-risk tumors (P = 0.07 for interaction). Adverse events within 30 days after surgery occurred in 21.4% of men, including one death.
41 In both the SPCG-4 and PIVOT, Men Who Underwent Surgery Experienced Significant Side-Effects SPCG-4 PIVOT
42 Prostate Cancer the Current Challenge
43 Who is at Risk? Vast majority of cases in men over 50 Risk continues to rise with increasing age Incidence and mortality vary with race/ethnicity Incidence 1.4 times greater in African American men, and death rates twice that in whites Five to 10% of prostate cancers may be inherited Affected first degree relative nearly doubles risk Risk increases with the number of affected first degree relatives A high-fat diet likely increases risk
44 New Directions in Screening (continued) There are three options: (1) USPSTF, AAFP: Discourage screening (2) ACS, et al.: Promote shared decision making Options 1 and 2 are status quo and more similar than they appear (3) Investigate risk based screening Avoid unnecessary PSA testing and unnecessary treatment Promote appropriate and more aggressive screening in men at high risk
45 PSA a current dilemma PSA is derided as a poor screening test due to an inability to distinguish between aggressive and indolent disease Data from the Prostate Cancer Prevention Trial (PCPT) showed there is no cut-off of PSA in which sensitivities and specificities are reasonably matched but rather a continuum of prostate cancer risk at all values of PSA. If one considers a biopsy Gleason score >7 as a measure of tumor aggressiveness, cut-off values of 4 and 2 ng/ml would miss 59.6% and 24.4% of such lesions, respectively. EUROPEAN UROLOGY 6 1 ( )
46 The real problem with PSA is how we re using it Two observations: PSA levels are very strongly correlated with clinically significant prostate cancer PSA has only modest diagnostic specificity and positive predictive value at commonly used thresholds for a positive finding. Routine screening with PSA leads to an excessive number of negative biopsies, and an excessive rate of diagnosis of indolent cancers How might we screen more effectively and in particular, put PSA to better use?
47 Rethinking screening, diagnosis, and treatment for prostate cancer Stop offering screening to men who are unlikely to benefit Men ages 70+ Men with life limiting co-morbidity (< 5 years) Screen men with low PSA ( 1ng/mL) less often than annually Possibly end screening at age 60 for men with persistently low PSA Consider alternative approaches to diagnosis and treatment decisions
48 PSA at or before age 50 as a predictor of advanced prostate cancer up to 25 years later BMC Medicine 2008, 6:6 In blood samples were obtained from a cohort of 21,277 men aged up to 50. Through 1999, 498 men were diagnosed with prostate cancer, and of these 161 had locally advanced or metastatic prostate cancers. Total PSA was a strong and statistically significant predictor of subsequent advanced cancer (area under the curve 0.791; p < ). Two-thirds of the advanced cancer cases occurred in men with the top 20% of PSA levels (0.9 ng/ml or higher).
49 Predicted probability of advanced prostate cancer by the total PSA level in anti-coagulated plasma measured at age Dashed lines indicate the 95% CIs.
50 Potential strategies to improve the predictive value of prostate cancer biopsy decisions If PSA thresholds have poor predictive value, are there other biomarkers that could improve the predictive value of decisions to biopsy? BMC Medicine 2008; 6:19 Vickers, et al. sought to determine whether a multivariable model including total, free, and intact PSA, and human kallikrein 2 (hk2)) could predict prostate biopsy outcome in previously unscreened men with elevated total PSA.
51 Potential strategies to improve the predictive value of prostate cancer biopsy decisions The study cohort comprised 740 men in Göteborg, Sweden, undergoing biopsy during the first round of the European Randomized study of Screening for Prostate Cancer Area-under-the-curve (AUC) for predicting prostate cancer at biopsy was calculated using a model that included: age and PSA (the 'laboratory' model) age, PSA and digital rectal exam (the 'clinical model) The laboratory and clinical models + the kallikreins. BMC Medicine 2008; 6:19
52 Biopsy Outcomes by PSA level, European Randomized Study of Prostate Cancer Screening in Göteborg, Sweden PSA cutoffs were 3 ng/ml For PSA range of , of 287 biopsies, only 3 high grade cancers were diagnosed At PSA ranges > 10, the yield is much higher (17-36%) BMC Medicine 2008; 6:19
53 Reduction in biopsies/cancers detected, as compared with the current study, using as a threshold for biopsy a 20% or higher probability of cancer Laboratory model = Age + PSA Clinical Model = Age + DRE + PSA * Using the clinical model, the addition of 4 additional makers (total + free + intact PSA and hk2) reduces biopsies by 60%, misses 17% of the low grade cancers, and 3% of high grade cancers BMC Medicine 2008; 6:19
54 Risk based strategies are influenced by prior testing Previous participation in PSA screening dramatically changed the performance of statistical models that were designed to predict biopsy outcome. Prior PSA Test No Prior PSA Test Vickers, et al. Cancer 2010;116:
55 Biopsies Conducted and Cancers Detected per 1000 Men With Elevated Prostate-Specific Antigen Who Had Recent Screening Using a Probability of Cancer 20% as the Threshold for Biopsy Vickers, et al. Cancer 2010;116:
56 Conclusion The long wait for end results from randomized trials has resulted in a kind of tunnel vision about how we think about prostate cancer screening This is likely to endure, since some embrace the results of the American trial, and some the European trial There is growing evidence that the harms associated with PSA testing can be measurably reduced
57 Conclusion What is needed today? Apart from not offering screening to adult men who have little chance of benefiting. Explore the potential for baseline testing (age?) to be a sentinel event Further explore the use of decision curve analysis to set threshold probabilities (10 40%) of prostate cancer for patient decision making Consider new approaches to decision making that would shift from the decision to screen to decisions about biopsy and treatment after screening
58 Thank You