To the Editor: The results of the Prostate, Lung, Colorectal,and Ovarian (PLCO) Cancer Screening Trial (ClinicalTrials.govnumber, NCT00002540
[ClinicalTrials.gov]
) by Andriole et al.1 were released prematurely,at the same time that Schröder et al.2 were reporting,on the basis of the European Randomized Study of Screening forProstate Cancer (ERSPC) (Current Controlled Trials number, ISRCTN49127736
[controlled-trials.com]
),that screening with prostate-specific antigen (PSA) saves lives(March 26 issue). The PLCO trial is being portrayed as the underpinningfor proposed changes in the mind-set of physicians and patientsthat prostate-cancer screening and treatment do more harm thangood. However, the PLCO trial used an outdated PSA cutoff andpermitted the enrollment of a large proportion of men (>40%)who had undergone prostate-cancer screening in the previous3 years. Furthermore, more than 50% of the control subjectsunderwent screening during the study. Most men with abnormalresults in the screening group did not undergo prostate biopsyfor 1 year or more. In the ERSPC trial, mortality curves didnot begin to diverge until 7 to 8 years. The PLCO investigatorsreported their results after a median follow-up of 5 to 6 yearsfor patients with cancer — 2 years before expected metastasesand 6 years before expected death from prostate cancer.3 ThePLCO trial was flawed at the starting gate, is a mere snapshottaken halfway around the track, and will never be informativeregarding the effect of screening on prostate-cancer mortalityamong healthy men who undergo intelligent screening and promptbiopsy and who receive effective treatment.
William J. Catalona, M.D. Northwestern University Feinberg School of Medicine Chicago, IL 60611 wcatalona{at}nmff.org
Dr. Catalona reports receiving research support from BeckmanCoulter, a manufacturer of PSA assays, and honoraria from GlaxoSmithKline.No other potential conflict of interest relevant to this letterwas reported.
References
Andriole GL, Crawford ED, Grubb RL III, et al. Mortality results from a randomized prostate-cancer screening trial. N Engl J Med 2009;360:1310-1319. [Erratum, N Engl J Med 2009;360:1797.] [Free Full Text]
Schröder FH, Hugosson J, Roobol MJ, et al. Screening and prostate-cancer mortality in a randomized European study. N Engl J Med 2009;360:1320-1328. [Free Full Text]
Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999;281:1591-1597. [Free Full Text]
To the Editor: It is important to recognize that the two randomizedtrials of PSA screening reported by Andriole et al. and Schröderet al. do not relate directly to the question of whether weshould be doing PSA screening. They relate to whether screeningshould be expanded beyond the levels that are typical of thepopulations in which the trials were conducted — levelsthat are reflected in the control groups and in pretrial conditions.
The U.S. National Health Interview Survey in 2005 showed that62.7% of men between the ages of 50 and 79 years who did nothave prostate cancer had undergone a PSA test and that 48.4%had been tested within the previous 2 years.1 These frequenciesare much greater than those in 14 economically developed countriesfor which similar data have been compiled. This difference maybe related to the fact that the decline in prostate-cancer mortalityhas been significantly faster in the United States than elsewhere.2A cautious interpretation of the results to date is that theUnited States would gain no mortality benefit by expanding PSAscreening beyond its already high levels, whereas European countrieswould benefit from such an expansion.
Samuel H. Preston, Ph.D. University of Pennsylvania Philadelphia, PA 19104 spreston{at}sas.upenn.edu
References
Ross LE, Berkowitz Z, Ekwueme DU. Use of the prostate-specific antigen test among U.S. men: findings from the 2005 National Health Interview Survey. Cancer Epidemiol Biomarkers Prev 2008;17;636-44.
Preston S, Ho JY. The US health care system and lagging life expectancy: a case study. Population Studies Center Working Paper 11. Philadelphia: University of Pennsylvania, February 2009.
To the Editor: The estimation by Andriole et al. that the PLCOtrial had a crossover contamination rate of 52% in the controlgroup is almost certainly an underestimate. Subjects were allowedto have one screening within 3 years before enrollment and anunlimited number of earlier screenings. Moreover, the estimationof contamination was made by surveys of patients, but men mayhave been screened without their knowledge during routine physicalexaminations.1 The strong effect of this contamination is reflectedin the fact that the likelihood of diagnosis was only 22% higherin the screening group than in the control group and that 94.3%of the tumors that were diagnosed in the control group wereat clinical stage I or II — tumors that are diagnosednearly exclusively by screening. Thus, the study was not a faircomparison between screening and no screening; instead, it wasa comparison between annual and ad hoc screening.
We agree entirely that overtreatment of patients with low-riskprostate cancer is a major public health problem.2 However,high-risk prostate cancer remains a leading cause of death fromcancer, and men should not be denied the opportunity for earlydiagnosis and selective treatment on the basis of data fromthe PLCO trial.
Matthew R. Cooperberg, M.D., M.P.H. Peter R. Carroll, M.D.,M.P.H. University of California, San Francisco San Francisco, CA 94143 pcarroll{at}urology.ucsf.edu
References
Federman DG, Goyal S, Kamina A, Peduzzi P, Concato J. Informed consent for PSA screening: does it happen? Eff Clin Pract 1999;2:152-157. [Medline]
Cooperberg MR, Broering JM, Kantoff PW, Carroll PR. Contemporary trends in low risk prostate cancer: risk assessment and treatment. J Urol 2007;178:S14-S19. [CrossRef][Web of Science][Medline]
To the Editor: The findings of the PLCO trial could be explainedby attribution bias, a misclassification that occurs when deathcertificates are used to determine the cause of death in a screenedpopulation.1 Improved prostate-cancer detection was attributedto screening on the basis of a 22% higher incidence of prostatecancer after 7 years in the screening group than in the controlgroup. The cause of death in cases detected by screening isoften mislabeled on death certificates as prostate cancer, despitethe greater probability of death from other causes among menwith prostate cancer.1 In contrast, prostate cancer would nothave been labeled as the cause of death among the undiagnosedcases in the control group, even if it had been the underlyingcause (i.e., unrecognized incidence-based mortality2). Therefore,misclassification of the cause of death because of overreportingin the screening group and underreporting in the control groupcould explain the observed lack of reduced mortality after screeningwith PSA testing and digital rectal examination.
Rohit P. Ojha, M.P.H. University of North Texas Health Science Center Fort Worth, TX 76107 rojha{at}hsc.unt.edu
Raymond Thertulien, M.D., Ph.D. ClinWorks Cancer Research Center Charlotte, NC 28207
Lori A. Fischbach, Ph.D., M.P.H. University of North TexasHealth Science Center Fort Worth, TX 76107
References
Feuer EJ, Merrill RM, Hankey BF. Cancer surveillance series: interpreting trends in prostate cancer -- part II: cause of death misclassification and the recent rise and fall in prostate cancer mortality. J Natl Cancer Inst 1999;91:1025-1032. [Free Full Text]
Incidence-based mortality. Bethesda, MD: National Cancer Institute, Statistical Research and Applications, 2006. (Accessed June 18, 2009, at http://srab.cancer.gov/statistics/ibm/.)
To the Editor: I was not able to discern whether the primary end-point statistic had crossed the efficacy boundary in theERSPC results. In his editorial, Barry1 conjectures that thedesign-adjusted statistical significance was based on a potentiallydata-driven cash-out on the remaining alpha at the third offour planned interim analyses. However, since Schröderet al. report that PSA-based screening reduced the rate of deathfrom prostate cancer by 20%, the boundary must have been crossed.If so, I think it is very important to point out that the reasonfor early reporting was the crossing of an efficacy boundary.After all, any attachment of meaning to the qualification "statisticallysignificant" within the context of a sequentially monitoredtrial requires a crossing of the efficacy boundary. Furthermore,any future discussion of the primary end point of this trialshould refer to the numbers therein, even after further follow-up,since follow-up reports can be considered only secondarily tothe primary analysis.2,3,4
Grant Izmirlian, Ph.D. National Cancer Institute Bethesda, MD 20892-7354 izmirlig{at}mail.nih.gov
References
Barry M. Screening for prostate cancer -- the controversy that refuses to die. N Engl J Med 2009;360:1351-1354. [Free Full Text]
Jennison C, Turnbull BW. Group sequential methods with applications to clinical trials. Boca Raton, FL: Chapman & Hall/CRC, 1999.
Proschan MA, Lan KKG, Wittes JT. Statistical monitoring of clinical trials: a unified approach. New York: Springer-Verlag, 2006.
Emerson SS, Kittelson JM, Gillen DL. Frequentist evaluation of group sequential clinical trial designs. Stat Med 2007;26:5047-5080. [CrossRef][Web of Science][Medline]
To the Editor: The studies by Schröder et al. and Andrioleet al. have insufficient discriminatory power to detect an effectof prostate-cancer screening on mortality.1 Such a conclusionappears to be generally impossible for statistical and practicalreasons and clearly marks a limitation of evidence-based medicine.Results of underpowered trials are prone to chance, which alsoexplains why the outcomes of the ERSPC and PLCO trials are ambiguous.
Furthermore, the ERSPC findings depend on a small number ofcases. If there had been 11 more deaths from prostate cancerin the screening group, the statistical significance would havevanished. With an additional 51 deaths, any benefit of screeningwould have been lost. Given these numbers, it should be notedthat all results from Portugal (involving 17,000 subjects) wereexcluded after the study was concluded because investigators"were unable to provide the necessary data." Furthermore, resultsfrom only a predefined "core age group" were analyzed. However,this age group differs from the age group specified a prioriin the Current Controlled Trials register. This leads to anestimated additional exclusion of 10,000 subjects.
Hans-Hermann Dubben, Ph.D. University Medical Center 20246 Hamburg, Germany dubben{at}uke.uni-hamburg.de
Dr. Andriole and colleagues reply: Catalona states that theresults of the PLCO trial were released prematurely and thatthey provide "a mere snapshot taken halfway around the track."The published results were complete for 7 years. These resultswill not change and are important for elderly men and othermen with a limited life expectancy. Indeed, in the PLCO trial,more men had been followed for 7 to 10 years than in the ERSPCtrial. However, we agree that additional follow-up is needed,and it is ongoing. Also, the PLCO trial was conducted accordingto standard U.S. practice, with physicians and subjects in thetwo groups likely to use similar PSA levels to trigger the decisionfor and the timing of biopsy, and treatments were similar inthe two groups.
We agree with Preston that one interpretation of the PLCO trialis that additional annual screening with PSA and digital rectalexamination does not further lower prostate-cancer mortality,as compared with the screening and treatment that are alreadyoccurring in the United States. However, we are not convincedby Preston that other countries should follow the example ofthe United States in advocating PSA screening, because the U.S.trends in death from prostate cancer may be due to other factors,including improved treatment, rather than screening.1 In addition,differences in treatment in the ERSPC trial (as outlined inSection 7 in the Supplementary Appendix accompanying the onlineversion of the article, available at NEJM.org) may explain thereported differentials in screening-related mortality.
Cooperberg and Carroll seem willing to discount the lack ofmortality benefit we demonstrated between the organized screeningin the intervention group and the "ad hoc" screening in thecontrol group. If, as they infer, deaths from prostate cancerwere largely due to high-risk cancers, it seems very probablethat more such prostate cancers would have been detected inthe intervention group than in the control group. Moreover,for the majority of men with lower-risk prostate cancer forwhom the date of diagnosis was advanced by screening (many menin the screening group and some in the control group), therecan only be a detriment associated with side effects of treatment.Also, even with high levels of screening in both groups in ourstudy, at 10 years, with 67% follow-up, 174 men still died ofprostate cancer.
With respect to the comments of Ojha et al.: we took great carein the death-review process to avoid the attribution bias theydescribe. Our unpublished comparison between the results ofthe blind reviews conducted by the end-point verification teamand the death-certificate data showed that such bias did notoccur.
Gerald L. Andriole, M.D. Washington University School of Medicine St. Louis, MO 63110
Anthony B. Miller, M.B. University of Toronto Toronto, ON M5S 1A1, Canada
Christine D. Berg, M.D. National Cancer Institute Bethesda, MD 20892 bergc{at}mail.nih.gov
Dr. Schröder and colleagues reply: We agreed with Prestonthat the results of the ERSPC and PLCO trials can be interpretedas answers to the question of whether PSA screening should beexpanded beyond the ongoing levels of screening in the respectivepopulations. In this sense, unlike the situation in the UnitedStates, in Europe the formal introduction of PSA screening mightyield an additional effect on prostate-cancer mortality. Inaddition, adjustment for noncompliance indicates that amongmen who actually undergo screening for prostate cancer, a decreaseof 27% in the rate of death from this cancer can be expected.Since uncertainty about the adverse effects of screening remainshigh at this time, this knowledge will be of crucial importancefor the decision-making process for patients and their physicians.We doubt that the decrease in mortality from prostate cancerin the United States since 1993 can be attributed to screeningalone. Screening has a lead time of 10 years or longer, whereasthe extensive use of PSA testing started after 1991.1 In addition,two treatment regimens, which were used in practice long before1993, have been shown to reduce prostate-cancer mortality.2,3Furthermore, the frequent use of statins in the United Statesis likely to have contributed to the decrease in prostate-cancermortality.4 Clearly, the reasons for the decrease in prostate-cancermortality in the United States are multifactorial.
Izmirlian raises a question about whether the efficacy boundarywas crossed in the ERSPC trial. We found the described effectof screening by applying a predefined significance limit ofP<0.05, which was adjusted for sequential testing, as describedin reference 4 of our article. We believe that Izmirlian isaddressing two different terminologies describing the same procedure,since a predefined significance limit is identical to an efficacy-boundarycrossing. In addition, at the time of the reported interim analysisof the ERSPC data (including data up to December 31, 2006),the data collection up to December 31, 2008, was already faradvanced. The limitations indicated by Izmirlian mainly applyto later follow-up reports, when screening in the control groupmay have become common.
Dubben questions whether the ERSPC and PLCO trials had sufficientstatistical power. Our data satisfy the power calculation thatis presented in reference 4 of our article, with sensitivityanalysis taking into account both noncompliance and contamination.Statistical significance was observed on the basis of predefinedlevels. We believe that a reduction of at least 20 to 30% inprostate-cancer mortality will be of great importance for publichealth and for individual patients worldwide. Agreement on thecore age group was reached in 1994 and fixed by contracts in1996. Data in the Current Controlled Trials register (ISRCTN)refer to inclusion only and do not specify plans for monitoringand evaluation. Unfortunately, there were no data availablefrom Portugal that would have justified the inclusion of thosefindings.
Fritz H. Schröder, M.D. Monique Roobol, Ph.D. Erasmus Medical Center 3000 CA Rotterdam, the Netherlands secr.schroder{at}erasmusmc.nl
Sue Moss, Ph.D. Institute of Cancer Research Sutton SM25NG, United Kingdom
for the ERSPC Investigators
References
Draisma G, Boer R, Otto SJ, et al. Lead times and overdetection due to prostate-specific antigen screening: estimates from the European Randomized Study of Screening for Prostate Cancer. J Natl Cancer Inst 2003;95:868-878. [Free Full Text]
Bill-Axelson A, Holmberg L, Filén F, et al. Radical prostatectomy versus watchful waiting in localized prostate cancer: the Scandinavian Prostate Cancer Group-4 randomized trial. J Natl Cancer Inst 2008;100:1144-1154. [Free Full Text]
Bolla M, Collette L, Blank L, et al. Long-term results with immediate androgen suppression and external irradiation in patients with locally advanced prostate cancer (an EORTC study): a phase III randomised trial. Lancet 2002;360:103-106. [CrossRef][Web of Science][Medline]
Platz EA, Leitzmann MF, Visvanathan K, et al. Statin drugs and risk of advanced prostate cancer. J Natl Cancer Inst 2006;98:1819-1825. [Free Full Text]
Previous
