PSA Screening for Prostate Cancer: A Debate on Implications of the PLCO Trial

Apr 17, 2012


David L. Graham, MD
Carle Physicians Group

We live in the prostate-specific antigen (PSA) era. The explosion inuse of the PSA test as a screen for prostate cancer since the 1980s hasbrought us to the point that its use can nearly be defined asubiquitous. If it is not ubiquitous yet, many clinicians would arguethat it should be. Despite this, no consensus of recommendations forits use exists among the main oncology and public health organizations.

Part of the reason for this may lie in the expectations for screeningtests. Previous requirements for a screening test included a test thatwas easy to perform at a reasonable cost, and which enabled anintervention leading to an effect on the outcome related to thatdisease. Gold-standard outcomes for screening studies have generallybeen disease-specific survival.

The inherent difficulty in demonstrating these criteria in a screeningtest for prostate cancer is that we have never had to evaluate a screenfor a disease that autopsy prevalence studies would suggest, dependingon the age group being screened, is present in 40% to 60% of thescreened patients. In addition, the relatively low lethality rate ofprostate cancer—14% by summarized American Cancer Societydata of 2007–2011—makes the impact of screeningpotentially more difficult to determine.

Large studies have demonstrated conflicting data regarding the effectof PSA screening. Updated data on the impact of PSA screening onprostate cancer-specific mortality from the National Cancer InstituteProstate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trialhave been recently published. In this column, Dr. William Catalona andDr. Derek Raghavan discuss the merits and critiques of the study andpresent their opinion on screening strategies for prostate cancer.

Dr.Graham is the Medical Director for Multispecialty Clinical ResearchServices at Carle Physicians Group, in Urbana, Illinois, and currentlyserves on ASCO’s Integrated Media and Technology Committee.Follow Dr. Graham on Twitter @davidgrahammd and read his commentary on

No Level 1 Evidence Indicates PSA Screening Improves Overall Survival

Derek Raghavan, MD, PhD, FACP
Levine Cancer Institute

The very sad aspect of the debate on PSA and screening for prostatecancer is that it seems to epitomize the epithet, “Oftenwrong, but never in doubt.” The discussion has repeatedlybeen mis-focused in current medical literature on whether the variousPSA screening studies were well-executed or flawed, and this has led tothe conclusion that the data published somehow misrepresent the utilityof PSA as a screening tool. These are not issues that help us tounderstand the situation.

Principlesof effective screening and analysis of trials
Before focusing on the specifics of prostate cancer, it is important tounderstand several key points that relate to any program of cancerscreening or cancer survival analysis:

  • For any cancer-screeningprogram to be useful, it must (by definition) improve overall survival,not just disease-specific survival.1 Another potentially useful rolefor cancer-screening programs is the reduction of the morbidity oftreatment. These well-tested and established criteria have frequentlybeen met for other cancers—specifically, mammographyin breast cancer, colonoscopy for colorectal cancer, and Pap smears forcervical cancer, among others. In a parallel situation, many years ago,annual chest x-rays were tested as a means of improving survival forlung cancer, failed the test, and were dropped from clinical practice.There is no reason why the bar should be lowered specifically tovalidate PSA as a screening test when the only improvement in survivalhas been tumor-specific survival, and no study has shown an overallsurvival benefit.
  • It is not appropriate todesign a clinical trial and then to negate the trial because itdoesn’t show the expected results. Even if a randomized trialhas design flaws, the presence of those flaws does not prove theopposite result from that demonstrated by the trial.
  • If a screening studydemonstrates that screening improves overall survival, that observationdoes not necessarily imply that all of the improvement in survival isdue to the test—in other tumors (and in prostate cancer),changes in diagnostic and staging techniques, introduction of adjuvanttherapies (e.g., breast cancer), changes in health practices (e.g.,smoking cessation vis-à-vis lung cancer), salvage therapies,novel chemotherapy and targeted therapeutics, and even improvedpalliative care can contribute to improved survival. Furthermore, inany clinical population, health-seeking behavior is often associatedwith improved outcomes—thus, the typical patient who seeksscreening tests also practices better dietary habits, exercises, andseeks medical attention for other conditions more promptly.
  • In the evaluation of anyrandomized clinical trial, focusing only on disease-specific survivalcan introduce a dangerous bias of interpretation. An example is theVeterans Administration study from 40 years ago, in which patients withadvanced prostate cancer were treated with high-dose estrogens in anattempt to gain control of their disease while a second group wasrandomly allocated to treatment with placebo.2 There was clear evidenceof improved cancer-specific survival in the estrogen-treated group, butoverall survival was identical; when the study was analyzed in detail,it became clear that high-dose estrogens reduced deaths from prostatecancer in the short term, but this reduction was offset by an increasein deaths from cerebrovascular and cardiovascular disease, created bythe therapeutic effect of high-dose estrogens. As a consequence, thismodality was dropped from our therapeutic armamentarium.
  • By definition, using a test for a symptomatic person (e.g., a man withsymptoms of prostatism) is not screening; screening is the use of atest to detect an illness in a patient with no symptoms or signs ofthat disease. Thus, the many clinical anecdotes about life-savinginterventions in men presenting with nonspecific urological symptomsare not relevant to a discussion of screening of asymptomaticpopulations.

Shouldwe be screening asymptomatic white males for prostate cancer using PSA?
Many randomized trials have addressed this question. Unfortunately, inall of these studies, there is a paucity of information regarding blackmen and men with positive family histories of prostate cancer, and wesimply don’t have enough information to answer the questionfor these population groups.

Given that outcomes (adjusted for age and socioeconomic status) in thisgroup are inferior to the rest of the community, my own view is that wemust emphasize education and provide structured approaches to screeningblack males, and any males with a positive family history, until clearinformation is available to resolve the issue.

However, much more information is available regarding prostatescreening in asymptomatic white males. Despite the significant designand execution flaws enumerated by Dr. Catalona, which should not beignored, the PLCO trial did not demonstrate an overall survival benefitfrom PSA-driven prostate screening, even after 13 years of follow up.3Randomized trials of screening for other cancers, by their nature,suffer from irregularities of follow-up and confounding factors (e.g.,crossover, prior testing, and failure to adhere to the plannedprotocol); however, as noted above, in many tumors, these studies haveshown an overall survival benefit from screening.

The two key European studies, ERSPC4 and Goteborg,5 have bothdemonstrated tumor-specific survival benefits, but no improvement inoverall survival. While one could argue that it is too early forsurvival benefit to have been demonstrated in prostate cancer (adisease with a long, natural history often occurring in elderly males)in any of these studies, I view that as exactly the reason why routinePSA screening should not be introduced into the general population.

Analyzingthe statistics
Many have claimed that there has been a reduction in cause-specificdeaths from prostate cancer since the introduction of PSA screening.I’m not sure that is true. For example, the U.S. governmentSEER statistics include the disclaimer, “includes onlyinvasive cancers unless otherwise specified.” In other words,more than 50% of the cases (those that are less aggressive) identifiedfrom screening activities are being omitted, which stacks the deck infavor of routine PSA screening.

By contrast, using total patient numbers generated from respectedAmerican Cancer Society publications, the figures are quite different.In 1981, in the United States, there were 20,790 deaths from prostatecancer6; in 1989, 25,943 deaths7; in 2012, it is projected that therewill be 28,170 deaths.8 This is hardly the pattern of major progressfrom a screening activity! The U.S. population has not swelled thatquickly, but the denominator of cases of low-risk disease, identifiedby random and enthusiastic PSA testing, certainly has grown quickly. Ifthe PSA believers really want to improve their data, why not add benignprostatic hyperplasia (BPH) to their denominator and improve theirrisk-benefit ratio even more?

At present, there is a cacophony of noise from the advocates of routinePSA screening, acknowledging that perhaps we overtreat men withprostate cancer, and noting the flaws in the extant PSA tools, leadingto their conclusion that we should now use risk-adapted, age-adapted,PSA-density, or velocity-adapted algorithms of screening. They may wellhave the right idea, and perhaps that is exactly the correct approachto developing a strategy to save lives. However, as in all otherdomains of medicine, and particularly those that potentially will costbillions of dollars to the community that is struggling to rationalizehealth expenditure, I challenge them to prove their hypotheses, and todo so this time with well-designed and well-executed clinical trials.

We should educate our population about the entity of prostate cancer,emphasizing symptoms, risk factors, and available treatments, defineprograms for black men and men with positive family histories, and takea much more rational approach to addressing this vexed problem, absentrhetoric and hype.

Dr.Raghavan is President of Levine Cancer Institute, CarolinasHealthCare System, and a Professor of Medicine at the University ofNorth Carolina. He has served on ASCO task forces related to cost ofcare, access to care, and health disparities, and received an ASCOStatesman Award in 2010.

1. Kramer BS, Croswell JM. Annu Rev Med.2009;60:125–37.
2. Blackard CE. Cancer Chemother Rep.1975;59:225–7.
3. Andriole GL, Crawford ED, Grubb RL 3rd, et al. J Natl Cancer Inst.2012;104:125–32.
4. Schröder FH, Hugosson J, Roobol MJ, et al. N Engl J Med.2009;360:1320–8.
5. Hugosson J, Carlsson S, Aus G, et al. Lancet Oncol.2010;11:725–32.
6. Silverberg E. CA Cancer J Clin.1981;31:13–28.
7. Silverberg E, Lubera JA. CA Cancer J Clin.1989;39:3–20.
8. Siegel R, Naishadham D, Jemal A. CA Cancer J Clin.2012;62:10–29.

What’s Wrong with PLCO: Drawbacks and Implications for PSA Screening

William J. Catalona, MD
Northwestern University Feinberg School of Medicine

Overallmortality: Not the primary endpoint
Dr. Raghavan’s focus on overall mortality and his example ofadministering toxic doses of estrogens to patients with advancedprostate cancer is more relevant to treatment trials than to screeningtrials. As with other cancer-screening programs, a key feature of theprostate cancer debate is reduction in cancer-specific mortality.Although overall mortality is an important outcome, the adoptions ofother larger cancer programs (e.g., breast and colon cancer) were basedon reductions of cancer-specific mortality. The primary endpoint of therandomized clinical trials of prostate cancer screening was prostatecancer-specific mortality (PCSM), not overall mortality. These trialscannot address the overall mortality issue at early follow-up. Overallmortality has a much lower statistical power in screening trials due tostatistical noise from deaths due to other causes. In acceptedscreening programs for other tumors, screening has not been shown tomodify overall mortality. Dr. Raghavan attributes the observedage-adjusted decreases in PCSM to causes other than PSA testing;however, there is simply no other mechanism that can reasonably accountfor the dramatic drop in age-adjusted PCSM since the introduction ofPSA.

ThePLCO trial
The design of the PLCO trial was flawed from the beginning, and itsfindings mislead and misinform. As a result, the popular and medicalmedia have reported faulty information, the public is confused, andphysicians are in the awkward position of having discussions withpatients based upon unfounded recommendations.

Enumeratingthe flaws
PLCO1–3 has by far the most flaws of the large, randomizedcontrolled trials of PSA-based screening for prostate cancer.

  • At least 43% of participantshad undergone prior screening before they enrolled in PLCO,1 and 4% to5% had undergone biopsy. This removed many high-risk tumors from thestudy population. This is reflected in the modest differences inhigh-grade and advanced-stage prostate cancer between the screening andcontrol arms that are substantially less than those observed in ERSPC4or Goteborg.5 Thus, in PLCO, PCSM events are fewer and occur later thanthey would without prescreening.
  • The most important flaw inPLCO is that at least 52% of controls were screened during the trial(“contamination”). This may, in fact, be anunderestimate, because PLCO investigators estimated contamination basedon surveys sent every one to two years to random 1% subsets ofcontrols. Controls also may have been screened, biopsied, and treatedduring the follow-up phase of the trial. Therefore, there was no purecontrol group, and many controls benefitted from“opportunistic” screening. Consequently, whileERSPC (≈15% contamination; 20% PCSM benefit) and Goteborg(≈3% contamination; 44% PCSM benefit) compared activescreening to minimal opportunistic screening; PLCO compared formalizedscreening in 85% to opportunistic screening in ≥52%. This highcontamination rate reduced differences in PCSM between the study arms.As there was no organized screening in PLCO after the screening phase,the two arms become even more similar, further decreasing differencesthat would have occurred with continued longitudinal screening. At theoutset, PLCO powered the study to demonstrate a 20% reduction in PCSMwith screening. The study investigators planned for non-compliance intheir power calculations, expecting 90% compliance in the screening armand 20% contamination in the control arm, but got 86% and 52%,respectively. To maintain the same level of power would require twicethe benefit anticipated to reach statistical significance. Thus, PLCOcannot be informative about whether longitudinal PSA-based screeningreduces PCSM. PLCO should not be interpreted as a “negativestudy” showing that screening does not work; it simply isunable to say anything about whether screening significantly reducesPCSM.
  • PLCO’s protocol(screening for six years with PSA using a >4 ng/mL cutoff plusdigital rectal examination for four years) is not representative ofcurrent screening methods. Prostate cancer risk-assessment tools nowwidely used for decision-making, such as baseline PSA, PSA velocity,percent-free PSA, etc., were not used in PLCO, and, therefore, theresults are not relevant to present-day screening.6
  • PLCO did not require menwith abnormal screening results to undergo biopsy; rather, the men andtheir physicians decided what further evaluation, if any, would occur.Only ≈40% with abnormal results underwent biopsy within oneyear, which compromised early detection and treatment.
  • PLCO included men up to age74. Since death from prostate cancer often occurs >13 yearsafter diagnosis, older men have more competing risks and are lesslikely to demonstrate a PCSM benefit. In Goteborg, with their shorterscreening interval and lower PSA cutoffs, less contamination and longerfollow-up, PCSM benefits occurred largely in younger men. This was alsothe case in ERSPC.
  • PLCO’s results notonly conflict with ERSPC and Goteborg but also are internallyinconsistent. PLCO reported a higher PCSM in the screening arm, butthat rate was 25% lower in men who had undergone ≥2 PSA tests atbaseline than in those not tested. In a subset analysis, screenedsubjects with minimal comorbidity had a 44% lower PCSM.2
  • PLCO recently updated themortality results (up to 13 years in 57%).3 The screening arm had ahigher prostate-cancer detection rate but a lower rate of advanced andhigh-grade tumors. Nevertheless, despite its more favorable tumorfeatures, the screening arm still had a 9% higher PCSM.
  • PLCO also re-examinedsubsets of their participants grouped by age, comorbidity, and pretrialPSA testing.3 On their second subset analysis, they changed thecomprehensive Charlson Index they used in their first analysis in favorof a more restrictive index that excluded patients with congestiveheart failure, peripheral vascular disease, connective-tissue disease,hemiplegia, HIV, renal disease, and dementia. Using the new definition,they reported no PCSM benefit in any of the subgroups; nevertheless,using the comprehensive definition used previously, there remained asignificant 27% lower PCSM in screened subjects with minimalcomorbidity. Regardless, the study was not powered to examine thisissue, and post hoc subset analyses should be interpreted with caution.
  • PLCO investigators speculatethat the PCSM benefits of ERSPC were due to differences in thetreatments offered to patients in the screening versus the control arms(i.e., screening-arm patients were more likely to receive radicalprostatectomy). Their premise is that treatment at a universityhospital is always better, and in at least one ERSPC center, patientswere more frequently treated at a university. Because screen-detectedprostate cancer is generally diagnosed earlier, it is more frequentlyamenable to radical prostatectomy. In ERSPC, the choice of therapy wasaffected more by other factors than by trial arm, with the trial armhaving only a minor role in treatment choice (e.g., χ2 of 8 ofa total χ2 of 2,428).7 PSA level, age, and tumor stage were themost important factors for treatment choice, and the study arm was notassociated with treatment after correction for age, PSA level, andtumor characteristics. Moreover, in Goteborg (having the greatest PCSMbenefit), men with localized disease in the two study arms receivedsimilar treatment. PLCO has not reported an analysis of their patientstreated in the community versus academic centers or adjusted forsurgical volume; thus, it is unknown how treatment differences affectedPLCO results.

Benefitsof PSA screening
During the PSA era, the United States has seen a >75% reductionin the rate of advanced prostate cancer at diagnosis and a >42%reduction of age-adjusted PCSM (Figure 1),8 with similar patternsoccurring in other countries where PSA screening is widespread.9Screening has not been adequately studied in high-risk populations suchas black men (only 4% of PLCO subjects were black), who have 50% higherprostate-cancer incidence and 200% higher PCSM, and whose outcomes haveimproved during the PSA era.7 Statistical modeling teams estimated 45%to 70% of this benefit is directly attributable to PSA screening.10

ERSPC and Goteborg provide valid level 1 evidence that PSA screeningreduces PCSM, whereas PLCO is noninformative on this issue. In acomprehensive evaluation of randomized controlled trials on prostatecancer screening, Allan et al. concluded, “The best evidencedemonstrates prostate cancer screening will reduce prostate cancermortality. It is time for the debate to move beyond thisissue...”11

Many men not wanting to suffer or die from prostate cancer would preferto accept the risks of screening to have the opportunity for earlydiagnosis when the tumor is still curable. Nearly all urologists andmost internists support PSA screening of appropriately selected men.12At present, PSA is the best test available for early diagnosis, and, infact, no other way to identify prostate cancer in its curable stages isnow available. To deny the opportunity for informed decision-makingneedlessly places men in harm’s way, and it would beunconscionable to implement that denial based upon flawed science.

Dr.Catalona, a surgical oncologist and prostate cancer specialist, isa professor in the Department of Urology at Northwestern FeinbergSchool of Medicine and Director of the Clinical Prostate Cancer Programat the Robert H. Lurie Comprehensive Cancer Center.

1. Andriole GL, Crawford ED, Grubb RL 3rd, et al. N Engl J Med.2009;360:1310–9.
2. Crawford ED, Grubb R 3rd, Black A, et al. J Clin Oncol.2011;29:355–61.
3. Andriole GL, Crawford ED, Grubb RL 3rd, et al. J Natl Cancer Inst.2012;104:125–32.
4. Schröder FH, Hugosson J, Roobol MJ, et al. N Engl J Med.2009;360:1320–8.
5. Hugosson J, Carlsson S, Aus G, et al. Lancet Oncol.2010;11:725–32.
6. Loeb S, Carter HB, Catalona WJ, et al. Eur Urol.2012;61:1–7.
7. Wolters T, Roobol JM, Steyerberg EW, et al. Int J Cancer.2010;126:2387–93.
8. Age-adjusted U.S. mortality rates by cancer site: all ages, allraces, male, prostate 1992–2008. Mortality source: USMortality Files, National Center for Health Statistics, CDC. Accessed 4February 2012.
9. Bouchardy C, Fioretta G, Rapiti E, et al. Int J Cancer.2008;123:421–9.
10. Etzioni R, Tsodikov A, Mariotto A, et al. Cancer Causes Control.2008;19:175–81.
11. Allan GM, Chetner MP, Donnelly BJ, et al. Canad Urol Assoc J.2011;5:416–21.
12. Sternberg S. US News Health. 28 October 2011. 23 February 2012.

Back to Top