SERUM PROSTATE-SPECIFIC ANTIGEN CONCENTRATION ISA POWERFUL PREDICTOR OF ACUTE URINARY RETENTION

AND NEED FOR SURGERY IN MEN WITH CLINICALBENIGN PROSTATIC HYPERPLASIA

CLAUS G. ROEHRBORN, JOHN D. MCCONNELL, MICHAEL LIEBER, STEVEN KAPLAN,JACK GELLER, GHOLEM H. MALEK, RONALD CASTELLANOS, SCOTT COFFIELD,

BRIAN SALTZMAN, MARTIN RESNICK, THOMAS J. COOK, AND JOANNE WALDSTREICHER,FOR THE PLESS STUDY GROUP

ABSTRACTObjectives. Prostate-specific antigen (PSA) is produced exclusively in the prostate gland and is currently themost useful clinical marker for the detection of prostate cancer. In this report, we examine whether serumPSA is also a predictor of important benign prostatic hyperplasia (BPH)-related outcomes, acute urinaryretention (AUR), and the need for BPH-related surgery.Methods. Three thousand forty men were treated with either placebo or finasteride in a double-blind,randomized study of 4-year duration. Serum PSA was measured at baseline, and baseline prostate volumewas measured in a 10% subset of 312 men. Probabilities and cumulative incidences of AUR and BPH-relatedsurgery, as well as reduction in risk of events with finasteride, were calculated for the entire patientpopulation, stratified by treatment assignment, baseline serum PSA, and prostate volume.Results. The risk of either needing BPH-related surgery or developing AUR ranged from 8.9% to 22.0%during the 4 years in placebo-treated patients stratified by increasing prostate volume and from 7.8% to19.9% when stratified by increasing serum PSA. In comparison with symptom scores, flow rates, andresidual urine volume, receiver operating characteristic curve analyses showed that serum PSA and prostatevolume were the most powerful predictors of spontaneous AUR in placebo-treated patients (area under thecurve 0.70 and 0.81, respectively). Finasteride treatment reduced the relative risk of needing surgery ordeveloping AUR by 50% to 74% and by 43% to 60% when stratified by increasing prostate volume andserum PSA, respectively.Conclusions. Serum PSA and prostate volume are powerful predictors of the risk of AUR and the need forBPH-related surgery in men with BPH. Knowledge of baseline serum PSA and/or prostate volume are usefultools to aid physicians and decision makers in predicting the risk of BPH-related outcomes and choosingtherapy for BPH. UROLOGY 53: 473–480, 1999. © 1999, Elsevier Science Inc. All rights reserved.

Benign prostatic hyperplasia (BPH) is a verycommon condition in elderly men. Histologic

evidence in the form of both stromal and glandu-

lar-epithelial hyperplasia can be found in approxi-mately 60% of men in their 60s, and 80% of men intheir 80s. Although not all these men require treat-

A complete list of the members of the PLESS Study Group is givenin the Appendix.

Dr. Waldstreicher and Thomas J. Cook are employees of Merck& Co., Inc., the sponsor of the study.

From the Departments of Urology, University of Texas South-western Medical Center at Dallas, Dallas, Texas; Mayo Clinic,Rochester, Minnesota; Columbia Presbyterian College of Physiciansand Surgeons, New York, New York; Jackson Foundation, Madison,Wisconsin; Ft. Myers Study Center, Fort Myers, Florida; Scott &White Memorial Hospital, Temple, Texas; Beth Israel Deaconess

Medical Center, Boston, Massachusetts; Case Western Reserve Uni-versity, Cleveland, Ohio; Department of Medicine, Mercy Hospitaland Medical Center, San Diego, California; and Departments of Bio-statistics and Clinical Research, Endocrinology and Metabolism,Merck Research Laboratories, Rahway, New Jersey

Reprint requests: Claus G. Roehrborn, M.D., Department of Urol-ogy, University of Texas Southwestern Medical Center at Dallas,5323 Harry Hines Boulevard, J8-130, Dallas, TX 75235-9110

Submitted: October 13, 1998, accepted (with revisions): No-vember 10, 1998

RAPID COMMUNICATIONCME ARTICLE

© 1999, ELSEVIER SCIENCE INC. 0090-4295/99/$20.00ALL RIGHTS RESERVED PII S0090-4295(98)00654-2 473

ment, in many it will cause bothersome symptomsthat may interfere with activities of daily living andreduce their quality of life.1 Long-term conse-quences of the disease may include acute urinaryretention (AUR) and the need for surgery, as wellas urinary tract infection, bladder function deteri-oration, and rarely renal failure due to obstruc-tion.2–4 Although limited clinical data are availableregarding the incidence of AUR and the need forsurgery in men followed up longitudinally, it isworthwhile to review the evidence at hand.

Among 50-year-old men, the lifetime incidenceof surgical or medical intervention for BPH is esti-mated to be 35%.5 In a 3-year study comparingwatchful waiting to transurethral resection of theprostate (TURP), 24% of men with moderatesymptoms assigned to watchful waiting underwentsurgery.6

AUR is a painful condition characterized by theinability to initiate voiding and empty the bladder.It may occur spontaneously in men with BPH or beprecipitated by surgery, anesthesia, or ingestion ofmedications such as alpha-sympathomimetics andanticholinergics. AUR occurred in 3% of patientswith moderate symptoms of BPH in the watchfulwaiting versus TURP study.6 In a cohort study of500 men with BPH who were candidates for pros-tate surgery and followed up for 4 years, the inci-dence of AUR was 25 per 1000 (40 per 1574) per-son-years.2 In the community-based longitudinalOlmsted County Study, the incidence of AUR in2115 men followed up for 4 years was 6.8 per 1000(57 per 8344) person-years.3

Although not very common, both AUR and theneed for surgery present a dramatic event in thenatural history of BPH for the patients affected bythem. It would therefore be advantageous to beable to predict in some form whether a given pa-tient is at higher or lower risk of experiencing ei-ther one of these outcomes.

Recent evidence suggests that prostate volumecorrelates with the risk of AUR in the OlmstedCounty community-based study.3 Prostate volumeis also a powerful predictor of symptom and flowrate improvement in men with BPH treated withthe 5-alpha-reductase inhibitor finasteride, whichdecreases conversion of testosterone to the morepotent androgen dihydrotestosterone, and de-creases prostate volume by approximately 20%.7

Treatment with finasteride also decrases the risk ofdeveloping AUR and needing BPH-related sur-gery.8 Although it would be intuitively reasonableto select patients at risk of therapy with finasterideon the basis of prostate volume, the digital rectalexamination (DRE) tends to underestimate pros-tate size considerably.9 However, a strong positivecorrelation between prostate volume and serum

PSA has been described,10 suggesting that PSA maybe useful as a proxy for prostate volume.

In the present analysis, we sought to determinewhether serum PSA, as a proxy for prostate vol-ume, is a useful predictor of long-term outcomes ofBPH, such as AUR and BPH-related surgery, in pa-tients followed up longitudinally with and withouttherapy with finasteride.

MATERIAL AND METHODS

SUBJECTSThree thousand forty men with clinical BPH diagnosed on

the basis of moderate-to-severe symptoms, a decreased peakurinary flow rate (less than 15 mL/s with a voided volume of150 mL or more; Urodyn 1000, Dantec, Mahway, NJ), and anenlarged prostate gland by DRE were enrolled in a 4-yearstudy comparing finasteride with placebo. Men receiving al-pha-blocking agents or antiandrogens and men with a historyof chronic prostatitis, recurrent urinary tract infections, pros-tate or bladder cancer or surgery, or a serum PSA greater than10 ng/mL were excluded. Men with serum PSA concentrationsbetween 4.0 and 9.9 ng/mL had to have a negative prostatebiopsy before enrollment.

STUDY DESIGNThe study was approved by the institutional review boards

at all 95 participating centers, and all men gave written in-formed consent. After a 1-month single-blind placebo lead-in,men were randomly assigned to receive placebo or 5 mg offinasteride (Proscar, Merck and Co., Inc., West Point, Pa)daily. Symptoms, bothersomeness, adverse events, and uri-nary flow rates were assessed every 4 months.8 Serum PSA wasmeasured every 4 months in the first year and then every 8months at a central laboratory. Physical examination and rou-tine hematologic and serum chemistry tests were performedyearly. Magnetic resonance imaging was performed at baselineand then yearly in a subset of participants at 13 centers (n 5312). All magnetic resonance images were read by a centralradiologist unaware of the treatment allocation and time ofimaging (ie, base line or follow-up). Additional details onstudy design have previously been described.8

AUR and surgery for BPH were predefined secondary endpoints. An Endpoint Committee, whose members were un-aware of the treatment assignment, reviewed all documenta-tion relating to episodes of AUR and all prostate surgeries forBPH, excluding surgery for prostate cancer. The EndpointCommittee classified episodes of AUR as spontaneous versusprecipitated (when contributing factors such as a cerebrovas-cular accident, urinary tract infection, surgery, anesthesia, andingestion of alpha-sympathomimetic drugs or anticholin-ergics were identified).

Complete data on outcomes, including 4-year follow-upinformation for men who had discontinued treatment, wereavailable for 92% of the men randomized. In the other 8%,complete information was available until discontinuation ofthe medication or up to the 6-month follow-up assessmentafter discontinuation.

STATISTICAL ANALYSISThe effect of baseline prostate volume and serum PSA on the

risk of a BPH-related outcome (developing AUR and need forBPH-related surgery) were assessed by dividing patients intotertiles of baseline prostate volume and baseline serum PSAand calculating the risk of developing an outcome by life-table, time-to-first event analysis, and Fisher’s exact test for

474 UROLOGY 53 (3), 1999

cumulative incidence. The finasteride-related reduction inrisk during 4 years was calculated using log-rank analyses.Comparisons among tertiles in event rates and treatment bytertile interactions were assessed using the Cox proportionalhazard test. All statistical tests were two-sided with an alpha of0.05 and a P value of less than 0.05 accepted as significant.

The measurement characteristics of baseline PSA and pros-tate volume in the prediction of BPH-related outcomes wereevaluated using receiver operating characteristic (ROC)curves. The area under the ROC curve (AUC) for a givenpopulation is given by the probability that a randomly chosenindividual in the affected population has a higher value of theindex (in this case PSA or prostate volume) than a randomlychosen individual in the nonaffected population. The AUCswere computed using the method of Hanley and McNeil.11

Differences between AUCs in the finasteride and placebogroup were tested using normal statistics with standard devi-ations computed by the same method.

RESULTS

At baseline, men assigned to finasteride and pla-cebo were similar in terms of age, demographics,symptom severity, peak flow rate, prostate volume,and serum PSA (Table I). Baseline characteristicsof the subset with prostate volume measurementswere similar to those in the entire study group.

The overall incidence of AUR was 7% with pla-cebo and 4% with finasteride (spontaneous AUR4% with placebo and 1% with finasteride; precipi-tated AUR 3% with placebo and 2% with finas-teride) and of BPH-related surgery, 10% in mentaking placebo and 5% in men taking finasteride.8

The incidence of AUR or surgery increased from8.9% to 22.0% from the first to the third tertile ofbaseline prostate volume for placebo-treated pa-tients; it remained relatively stable between 5.1%and 5.6% for finasteride-treated patients (Fig. 1A).As a result, the risk reduction with finasteride in-creased from 50% in the smallest to 74% in thelargest prostate volume tertile. When stratified by

baseline serum PSA, the risk increased from 7.8%to 19.9% for the placebo group (P , 0.001) andfrom 4.4% to 8.3% for the finasteride group (P 50.035), resulting in a finasteride-related reductionof risk from 43% in the lowest to 60% in the highesttertile of serum PSA (Fig. 1B).

FIGURE 1. Four-year incidences of either AUR or BPH-related surgery in patients treated with placebo or fin-asteride, stratified in tertiles by (A) baseline prostatevolume (subset of 10% of patients) or (B) baseline se-rum PSA. Arrows denote reduction in risk by the log-rank test. †One placebo patient had a prostate volumeof 222.

TABLE I. Baseline characteristics of men in the finasteride and placebo groups

Parameter

Placebo Finasteride

Mean 6 SD n Mean 6 SD n

Age (yr) 64 6 7 1503 64 6 6 1513Quasi-AUA symptom score* 15 6 6 1503 15 6 6 1513Peak flow rate (mL/s) 11 6 4 1196 11 6 4 1208Prostate volume (mL)† 55 6 26 155 54 6 25 157

First tertile (14–41) 32 6 6 45 33 6 6 59Second tertile (42–57) 49 6 5 60 47 6 5 44Third tertile (58–150‡) 81 6 29 50 82 6 22 54

Serum PSA (ng/mL) 2.8 6 2.1 1498 2.8 6 2.1 1512First tertile (0.2–1.3) 0.9 6 0.3 511 0.8 6 0.3 472Second tertile (1.4–3.2) 2.2 6 0.6 514 2.2 6 0.6 536Third tertile (3.3–12.0) 5.4 6 1.7 473 5.4 6 1.7 504

None of the differences was significant.* Quasi-AUA symptom score based on an adaptation of the American Urological Association (AUA) symptom score, also known as the IPSS (International Prostate SymptomScore).† Prostate volume was only measured in approximately 10% of patients in 13 of the participating centers.‡ One additional patient had a prostate volume of 222 mL.

UROLOGY 53 (3), 1999 475

Figure 2 displays Kaplan-Meier curves for the in-cidence of AUR or surgery for patients treated withplacebo or finasteride, stratified by baseline PSA.Patients in the lowest tertile of serum PSA (0.0 to1.3 ng/mL) had the lowest risk of either one of theevents, and the benefit of finasteride over placebo isminimal for the first 2 years. In this tertile, the over-all 4-year relative risk for finasteride- versus place-bo-treated patients is 0.57 (95% confidence interval[CI] 0.35 to 0.95), with a 43% risk reduction (P 50.030; Fig. 2A). For patients in the second tertile(serum PSA between 1.4 and 3.2 ng/mL), the 4-year

risk for finasteride- versus placebo-treated patientsis 0.54 (95% CI 0.37 to 0.80), with a 46% riskreduction with finasteride treatment (P 5 0.002;Fig. 2B). Finally, the relative risk for finasteride-versus placebo-treated patients in the third tertileis 0.40 (95% CI 0.29 to 0.56), with a 60% riskreduction with finasteride (P , 0.001; Fig. 2C).

The cumulative incidences for spontaneousAUR, all AUR (spontaneous and precipitated com-bined), and BPH-related surgery for 4 years by in-crements of baseline serum PSA thresholds areshown in Figure 3. Although in the placebo groupthe cumulative incidence for spontaneous (Fig.3A) and all AUR (Fig. 3B) increases with increasingserum PSA values (P , 0.001), in the finasteride-treated patients this effect is nearly absent. Thecumulative incidence of BPH-related surgery in-creases linearly across PSA values in placebo-treated patients from 10% to 24% (P , 0.001),whereas in finasteride-treated patients it increasesonly in men with a baseline PSA greater than 5.0ng/mL (P 5 NS) (Fig. 3C).

ROC curve analyses evaluating the performanceof baseline serum PSA in predicting outcomes incomparison to the more traditional baseline pa-rameters of BPH, including symptom severity,bothersomeness, peak urinary flow rate, residualurine volume, and age, are shown in Table II.Within the placebo group, serum PSA and prostatevolume were the best predictors for all AUR (AUCs0.68 and 0.69, respectively), as well as for sponta-neous AUR (AUC 0.70 and 0.81, respectively). In-terestingly, serum PSA, prostate volume, and peakflow rate had higher AUC values in the placebogroup for AUR than for BPH-related surgery.Symptom severity, bothersomeness scores, and re-sidual urine had higher AUC values in the placebogroup for BPH-related surgery than for AUR. Agehad similar AUC values in the placebo group forboth AUR and BPH-related surgery. In general, theAUCs were numerically higher for the placebogroup than for the finasteride group, suggestingthat finasteride treatment weakens the relationshipbetween baseline values and the occurrence ofBPH-related outcomes (Fig. 4). Neither symptomseverity nor bothersomeness of symptoms had anypredictive value (AUCs 0.49 and 0.47, respec-tively) for the development of AUR in finasteride-treated patients (Table II).

COMMENT

Serum PSA is currently the most widely usedmarker for prostate cancer detection, and a yearlymeasurement is recommended in men older than50 years to aid in the early detection of prostatecancer. The observation that there is a strong log-linear relationship between serum PSA and pros-

FIGURE 2. Kaplan-Meier curves for the risk of all AURor BPH-related surgery for patients treated with placeboor finasteride by baseline serum PSA stratified in tertiles.(A) Risk for the lowest tertile of serum PSA (0.2 to 1.3ng/mL), (B) risk for the second or middle tertile (1.4 to3.2 ng/mL), and (C) risk for the third or highest serumPSA tertile (3.3 to 12.0 ng/mL).

476 UROLOGY 53 (3), 1999

tate volume in men with BPH10 has led us to con-sider that PSA may also predict those men atincreased risk of developing AUR or needing BPH-related surgery.

In the longitudinal community-based OlmstedCounty study,3 it had been shown that men withprostate volumes greater than 30 mL by transrectalultrasound had almost four times the odds of ex-

periencing AUR compared with those with pros-tate volumes less than 30 mL. The tertile analysis(Fig. 1) presented here demonstrates a very strongrelationship between baseline prostate volume andserum PSA in predicting the incidence of BPH-re-lated surgery or AUR during 4 years. The significantincrease in the risk of experiencing these compli-cations with placebo treatment is nearly com-pletely obliterated with finasteride treatment. Thisphenomenon leads to a greater benefit of finas-teride over placebo in men with either larger pros-tate volumes or higher baseline PSA in avoidingthese complications. About 1 of 5 patients in thehighest tertiles are likely to experience either AURor surgery for BPH, and the risk reduction withfinasteride treatment is 74% (based on prostatevolume) and 60% (based on serum PSA).

A second important question for clinicians iswhether the risk increases over time of observa-tion. In fact, in placebo-treated patients, the cumu-lative risk increases in a linear fashion for all threeserum PSA tertiles (Fig. 2). With the exception ofthe lowest PSA tertile, where there is no apparentbenefit of finasteride over placebo in the first 2years (Fig. 2A), the difference in risk betweentreatment groups in the two higher tertiles (greaterthan 1.3 ng/mL) becomes evident as early as thefirst follow-up visit at 4 months (Figs. 2B and C).

Physicians may have different thresholds regard-ing their use of preventive healthcare measures.Figure 3 provides a graphic assessment to aid inthis decision. The cumulative incidence of sponta-neous AUR for placebo-treated patients increasesdramatically with serum PSA levels above approx-imately 1.3 ng/mL. In fact, although the cumula-tive risk for all patients is approximately 4% or 1 in25 men in 4 years, it reaches 9% or nearly 1 in 10patients for those with a PSA greater than 4.0ng/mL at baseline (Fig. 3A). At the same time, therisk remains unchanged for the entire serum PSAspectrum for finasteride-treated patients. Similarobservations hold true for the cumulative risk ofboth spontaneous and precipitated AUR and BPH-related surgery (Figs. 3B and C).

The ROC curve analyses show a trend toward areduction in the predictive power of many baselineparameters with finasteride therapy when com-pared with placebo. This is not unexpected as ther-apy with finasteride significantly interferes withthe natural history of the disease process. Forspontaneous AUR, the purest outcome in terms ofbeing the least likely to be influenced by either thepatient or physician, both prostate volume (AUC0.81) and serum PSA (AUC 0.70) are the mostpowerful predictors in the placebo-treated patients(Fig. 4C). That serum PSA and prostate volume areboth similarly significant predictors of outcomes is

FIGURE 3. Incidences of (A) spontaneous AUR, (B)both spontaneous and precipitated AUR, and (C) BPH-related surgery in placebo- and finasteride-treated pa-tients during 4 years by incremental baseline serumPSA thresholds.

UROLOGY 53 (3), 1999 477

likely a reflection of the strong correlation betweenthe two parameters.10

It is interesting to speculate why the symptomseverity and bothersomeness scores are better pre-dictors of BPH-related surgery than of AUR. Spon-taneous AUR is an event that cannot be influencedby the patient, the physician, or by their interac-tion. A placebo effect is most unlikely in such setting.Precipitated AUR, triggered by surgery, ingestionof drugs such as alpha-adrenergics or anticholin-ergics, is clearly a less reliable end point in thenatural evolution of BPH. The incidence of BPH-related surgery, on the other hand, can be substan-tially influenced by the interaction between patientand physician. It is likely that this patient-physi-cian interaction, the discussion of perceived pro-gression and improvement of the condition basedon scores, may have an influence on the need forsurgery. It is intuitively obvious that in the face ofrising symptom scores or decreasing peak flowrates, the physician may suggest that the patientconsider a surgical procedure. This subjective ele-ment is probably less prevalent in a clinical trialsetting, where both patients and physicians feelcompelled to continue the study until it ends.However, when the trial is scheduled for 4 years,clinical judgment may lead the physician to at leastdiscuss with the patient possible surgery, as other-wise the patient would have to live with worseningsymptoms for a considerable period.

That finasteride attenuates the predictive powerof prostate volume and serum PSA regarding sur-gery and AUR was anticipated, as it has beenclearly shown to change the natural history of the

disease by shrinking the prostate gland and pre-venting further measurable growth. In fact, tomake predictions about the risk of surgery and/orAUR once treatment is initiated is less importantthan to be able to give patients and healthcare pro-viders information before a treatment decision re-garding possible future BPH-related outcomes.

Several limitations of our study need to be recog-nized. First, even though it is the longest BPH trialever conducted, in the lifespan of a patient withBPH it still covers only a fraction of the entire du-ration during which the patient is at risk. Extrapo-lations of the data over extended periods have to beundertaken with great caution. Second, all patientsin this study had to have an enlarged prostate byDRE at baseline to be eligible for participation. De-spite the stratification by PSA levels and the strongclinical correlation between serum PSA and pros-tate volume, we cannot state with certainty thatsimilar results would be obtained in a cohort ofmen not selected for enlarged prostate glands.

Prostate volume and serum PSA are good candi-date parameters to aid in the individualized discus-sion that takes place between patients and physiciansbefore initiation of therapy for BPH. Finasteridedecreases the risk of developing a BPH-related out-come by approximately half in all subgroups exam-ined. However, the absolute risk of having an out-come is substantially different across the differentlevels of PSA and prostate volume. Risk is vieweddifferently by patients, physicians, and administra-tors. The data provided in this report should behelpful to all parties involved in the decision of

TABLE II. Area under the curve 6 standard error values for ROCcurves for several baseline parameters for spontaneous acute

urinary retention and prostate-related surgeryParameter Finasteride Placebo P Value*

Spontaneous AURSerum PSA 0.53 6 0.06 0.70 6 0.03 0.012Prostate volume 0.67 6 0.04 0.81 6 0.11 0.665Peak flow rate 0.67 6 0.06 0.62 6 0.04 0.510Residual urine volume 0.46 6 0.07 0.56 6 0.04 0.224Quasi-AUA symptom score 0.49 6 0.06 0.55 6 0.04 0.440Bothersomeness score 0.47 6 0.07 0.58 6 0.04 0.168Age 0.57 6 0.06 0.53 6 0.04 0.562

SurgerySerum PSA 0.59 6 0.03 0.62 6 0.02 0.461Prostate volume 0.49 6 0.09 0.63 6 0.08 0.213Peak flow rate 0.59 6 0.04 0.57 6 0.03 0.692Residual urine volume 0.52 6 0.03 0.60 6 0.02 0.046Quasi-AUA symptom score 0.60 6 0.03 0.59 6 0.02 0.761Bothersomeness score 0.55 6 0.03 0.60 6 0.02 0.197Age 0.60 6 0.03 0.57 6 0.03 0.507

KEY: ROC 5 receiver operating characteristic; AUR 5 acute urinary retention; PSA 5 prostate-specific antigen; Quasi-AUA 5 adaptation of American Urological Association symptom score.* Represents the between-group P value.

478 UROLOGY 53 (3), 1999

whether to treat BPH based on predictable risksand predictable reductions in risk with finasteride.

CONCLUSIONS

Serum PSA strongly correlates with prostate vol-ume in men with BPH, and both parameters predictequally well the risk of AUR and the need for BPH-related surgery. The incidence of both untowardoutcomes increases nearly linearly with increasingserum PSA and prostate volume in men treated

with placebo—or watched conservatively—and itis reduced by 50% or more in men on finasterideindependent of baseline serum PSA and prostatevolume. The predictable risk of retention and needfor surgery, and the reduction of risk with finas-teride, should help all parties involved in decisionmaking (patients, physicians, and administrators)to decide whether and how to treat symptomaticBPH.

REFERENCES1. Girman CJ, Epstein RS, Jacobsen SJ, et al: Natural his-

tory of prostatism: impact of urinary symptoms on quality oflife in 2115 randomly selected community men. Urology 44:825–831, 1994.

2. Barry MJ, Fowler FJ Jr, Bin L, et al: The natural historyof patients with benign prostatic hyperplasia as diagnosed byNorth American urologists. J Urol 157: 10–15, 1997.

3. Jacobsen SJ, Jacobson DJ, Girman CJ, et al: Natural his-tory of prostatism: risk factors for acute urinary retention.J Urol 158: 481–487, 1997.

4. Guess HA: Benign prostatic hyperplasia: antecedentsand natural history. Epidemiol Rev 14: 131–153, 1992.

5. Oesterling JE: Benign prostatic hyperplasia: a review ofits histogenesis and natural history. Prostate Suppl 6: 67–73,1996.

6. Wasson JH, Reda DJ, Bruskewitz RC, et al, for the Vet-erans Affairs Cooperative Study Group on Transurethral Re-section of the Prostate: A comparison of transurethral surgerywith watchful waiting for moderate symptoms of benign pros-tatic hyperplasia. N Engl J Med 332: 75–79, 1995.

7. Boyle P, Gould AL, and Roehrborn CG: Prostate vol-ume predicts outcome of treatment of benign prostatic hyper-plasia with finasteride: meta-analysis of randomized clinicaltrials. Urology 48: 398–405, 1996.

8. McConnell JD, Bruskewitz R, Walsh P, et al, for theFinasteride Long-Term Efficacy and Safety Study Group: Theeffect of finasteride on the risk of acute urinary retention andthe need for surgical treatment among men with benign pros-tatic hyperplasia (see comments). N Engl J Med 338: 557–563,1998.

9. Roehrborn CG, Girman CJ, Rhodes T, et al: Correlationbetween prostate size estimated by digital rectal examinationand measured by transrectal ultrasound. Urology 49: 548–557, 1997.

10. Roehrborn CG, Boyle P, Gould AL, et al: Serum pros-tate-specific antigen as a predictor of prostate volume in menwith benign prostatic hyperplasia. Urology 53: 581–589,1999.

11. Hanley JA, and McNeil BJ: The meaning and use of thearea under a receiver operating characteristic (ROC) curve.Radiology 143: 29–36, 1982.

APPENDIXThe PLESS Study Group includes (in alphabetical order):

A. Aigen, P. Albertsen, R. Anderson, G. Andriole, S. Auerbach,M. Bamberger, J. Bannow, W. Barzell, D. Bergner, J. Bonilla,R. B. Bracken, W. Brannan, W. Bremner, T. Brown, R. Bruske-witz, R. Castellanos, S. Childs, K. S. Coffield, T. Cook, C. Cox,E. D. Crawford, B. Dalkin, R. W. deVere White, G. Drach,H. Epstein, C. Ercole, D. Falcone, D. Finnerty, W. Fitch,M. Flanagan, J. Fowler, H. Fuselier, D. Garvin, J. Geller,R. Gibbons, P. Gilhooly, M. Gittelman, S. Glickman, J.Gottesman, T. Gray, J. Grayhack, H. Guess, L. Harrison, W.Hellstrom, R. Herlihy, G. B. Hodge, Jr., H. L. Holtgrewe, R.Huben, P. Hudson, C. L. Jackson, E. Johnson, D. Kadmon,

FIGURE 4. ROC curve analyses for the ability of base-line serum PSA to predict (A) the need for BPH-relatedsurgery, (B) both spontaneous or precipitated AUR, and(C) spontaneous AUR only for patients treated withplacebo or finasteride.

UROLOGY 53 (3), 1999 479

S. Kandzari, S. Kantor, S. Kaplan, M. Koppel, G. Kornitzer,D. Kozlowski, O. Kurzer, R. Labasky, J. Libertino, M.Lieber, R. Lund, S. Luttge, D. Lynch, G. Malek, N. Mangel-son, A. Matsumoto, J. D. McConnell, W. S. McDougal, A.Melman, D. Milam, R. Milsten, J. Mitchell, D. Mobley, P.Narayan, J. C. Nickel, L. Oppenheimer, F. Pappas, R. Parra,L. Peterson, J. Rajfer, P. Reddy, M. Resnick, O. F. Rigby,

C. G. Roehrborn, N. Romas, S. Rosenberg, S. Rosenblatt, S.Rous, C. Rowe, J. Roy, B. Saltzman, W. P. Sawyer, P. Schell-hammer, J. Schmidt, K. Short, T. Shown, D. Siegel, M. So-loway, T. Stanisec, B. Stein, E. Stoner, M. Sullivan, D. Suss-man, A. M. Taylor, L. Tenover, J. Waldstreicher, P. Walsh,D. Wang, F. Wei, S. Weiner, G. Wells, H. Wessells, C.White, H. Wise, and G. Zhang.

480 UROLOGY 53 (3), 1999