Section Editors Robert H Fletcher, MD, MSc Michael P O'Leary, MD, MPH

Transcription

1 1 de :24 Official reprint from UpToDate UpToDate Author Richard M Hoffman, MD, MPH Disclosures Section Editors Robert H Fletcher, MD, MSc Michael P O'Leary, MD, MPH Deputy Editor David M Rind, MD All topics are updated as new evidence becomes available and our peer review process is complete. Literature review current through: Apr This topic last updated: Abr 24, INTRODUCTION Prostate cancer is common and a frequent cause of cancer death. In the United States, prostate cancer is the most commonly diagnosed visceral cancer; in 2012, there were expected to be about 239,000 new prostate cancer diagnoses and about 30,000 prostate cancer deaths [1]. Prostate cancer is second only to nonmelanoma skin cancer and lung cancer as the leading cause of cancer and cancer death, respectively, in US men. Worldwide, in 2008 there were estimated to be 903,000 new cases of prostate cancer and 258,000 prostate cancer deaths making it the second most commonly diagnosed cancer in men and the sixth leading cause of male cancer death [2]. For an American male, the lifetime risk of developing prostate cancer is 16 percent, but the risk of dying of prostate cancer is only 2.9 percent [3]. Many more cases of prostate cancer do not become clinically evident, as indicated in autopsy series, where prostate cancer is detected in one-third of men under the age of 80 and in two-thirds of older men (see "Risk factors for prostate cancer", section on 'Age') [4]. These data suggest that prostate cancer often grows so slowly that most men die of other causes before the disease becomes clinically advanced. Prostate cancer survival is related to many factors, especially the extent of tumor at the time of diagnosis. The five-year relative survival among men with cancer confined to the prostate (localized) or with just regional spread is 100 percent compared with 31.9 percent among those diagnosed with distant metastases [3]. While men with advanced stage disease may benefit from palliative treatment, their tumors are generally not curable. Thus, a screening program that could identify asymptomatic men with aggressive localized tumors might be expected to substantially reduce prostate cancer morbidity, including urinary obstruction and painful metastases, and mortality. Prostate-specific antigen (PSA) testing revolutionized prostate cancer screening. Although PSA was originally introduced as a tumor marker to detect cancer recurrence or disease progression following treatment, it became widely adopted for cancer screening by the early 1990s. Subsequently, professional societies issued guidelines supporting prostate cancer screening with PSA [5,6]. PSA testing led to a dramatic increase in the incidence of prostate cancer, peaking in 1992 (figure 1) [7]. The majority of these newly-diagnosed cancers were clinically localized (figure 2), which led to an increase in radical prostatectomy and radiation therapy, aggressive treatments intended to cure these early-stage cancers [8-11]. However, prostate cancer screening has been a controversial issue because decisions were made about adopting PSA testing in the absence of efficacy data from randomized trials [12]. Subsequently, the European Randomized Study of Screening for Prostate Cancer (ERSPC) reported a small absolute survival benefit with PSA screening after nine years of follow-up [13]; however, 48 additional patients would need to be diagnosed with prostate cancer to prevent one prostate cancer death. Although the report did not address quality of life outcomes, considerable data show the potential harms from aggressive treatments, including erectile dysfunction, urinary incontinence, and bowel problems [14]. Further sustaining the uncertainty surrounding screening, a report from the large United States trial, the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, published concurrently with the European trial, found no benefit for annual PSA and digital rectal examination (DRE) screening after 7 to 10 years of follow-up [15].

2 2 de :24 This topic reviews the screening tests that are available for prostate cancer, the efficacy of screening, and the recommendations of major medical associations and societies regarding screening for prostate cancer. Risk factors and the clinical manifestations and diagnosis of prostate cancer are discussed separately. (See "Risk factors for prostate cancer" and "Clinical presentation and diagnosis of prostate cancer".) PROSTATE SPECIFIC ANTIGEN (PSA) PSA is a glycoprotein produced by prostate epithelial cells. PSA levels may be elevated in men with prostate cancer because PSA production is increased and because tissue barriers between the prostate gland lumen and the capillary are disrupted, releasing more PSA into the serum. (See "Measurement of prostate specific antigen".) Studies have estimated that PSA elevations can precede clinical disease by 5 to 10 years [16,17] or even longer [18]. However, PSA is also elevated in a number of benign conditions (table 1), particularly benign prostatic hyperplasia (BPH) and prostatitis. (See "Clinical manifestations and diagnostic evaluation of benign prostatic hyperplasia" and "Acute bacterial prostatitis".) Measuring PSA In addition to the PSA elevations seen with BPH, there are transient causes of PSA elevation (table 1), some of which are significant enough to affect the performance of PSA measurement as a screening test. We describe PSA values in ng/ml throughout this topic, but this is equivalent to the SI units of mcg/l; that is, 4 ng/ml = 4 mcg/l. (See "Measurement of prostate specific antigen", section on 'Causes of an elevated serum PSA'.) PSA has a half-life of 2.2 days [19], and levels elevated by different benign conditions will have variable recovery times [20-22]. PSA testing should be deferred accordingly: Digital rectal examination (DRE) has minimal effect on PSA levels, leading to transient elevations of only 0.26 to 0.4 ng/ml, and PSA can be measured immediately after DRE [23,24]. Ejaculation can increase PSA levels by up to 0.8 ng/ml, though levels return to normal within 48 hours [25,26]. We do not usually ask men to abstain from sexual activity prior to PSA measurement. However, if an initial measurement is high enough to potentially prompt an intervention (ie, biopsy), but close to a borderline value, it is appropriate to repeat the PSA measurement after having the man abstain from ejaculation for at least 48 hours. Bacterial prostatitis may elevate PSA levels [27], but they generally return to baseline six to eight weeks after symptoms resolve. Asymptomatic prostatic inflammation can also elevate PSA levels [28], but this diagnosis is made on biopsy and so cannot generally be used to defer screening tests [27]. Prostate biopsy may elevate PSA levels by a median of 7.9 ng/ml within 4 to 24 hours following the procedure [20]. Levels will remain elevated for two to four weeks. Similarly, a transurethral resection of the prostate (TURP) can elevate PSA levels by a median of 5.9 ng/ml [20]. Levels will remain elevated for a median time of approximately three weeks. A screening PSA test should not be performed for at least six weeks following either of these procedures. Acute urinary retention may elevate PSA levels, but the levels can be expected to decrease by 50 percent within one to two days following resolution. A screening PSA test should not be performed for at least two weeks following an episode of acute urinary retention. The five-alpha reductase inhibitors finasteride and dutasteride lower PSA levels. Finasteride lowers PSA by a median 50 percent within six months of use, though the effects can vary widely, ranging from 81 percent to +20 percent [29]; dutasteride has been reported to reduce PSA 48 to 57 percent [30]. Some experts recommend doubling the measured PSA value before interpreting the result for patients on finasteride [31]. Longitudinal results from the Prostate Cancer Prevention Trial suggest that PSA values be corrected by a factor of 2 for the first two years of finasteride therapy, and by 2.5 for longer-term use [32]. Test performance Determining the accuracy of PSA testing has been difficult because most men with normal PSA values will not undergo biopsy unless their digital examination is abnormal. This work-up bias tends to overestimate sensitivity and underestimate specificity [33]. Performance can also be overestimated because PSA often detects clinically-unimportant cancers. (See 'Overdiagnosis' below.)

3 3 de :24 Another problem in assessing the accuracy of PSA is that the transrectal needle biopsy is not a perfect gold standard. Investigators have suggested that the false-negative rate can range from 10 to 20 percent [34,35], though the recent trend towards obtaining 12 samples has increased the detection rate [36,37]. Additionally, protocols that use large numbers of biopsies to evaluate patients with an elevated PSA may be detecting incidental cancers that were not the etiology of the PSA elevation. One review that assumed that nonpalpable cancers smaller than 1.0 cm 3 would not cause elevated PSA levels estimated that approximately 25 percent of cancers detected by PSA screening were too small to have accounted for the PSA rise that prompted a biopsy [38]. The diagnostic performance of PSA ideally needs to be calibrated against clinically-important cancers. However, there is no consensus on defining such cancers. Although many experts consider tumors with Gleason scores 7 and volumes >0.5 cm 3 to have a greater risk for progression, there is no certainty that these cancers will lead to early death or reduce quality of life [39]. Sensitivity and specificity The traditional cutoff for an abnormal PSA level in the major screening studies has been 4.0 ng/ml [40-43]. The American Cancer Society systematically reviewed the literature assessing PSA performance [44]. In a pooled analysis, the estimated sensitivity of a PSA cutoff of 4.0 ng/ml was 21 percent for detecting any prostate cancer and 51 percent for detecting high-grade cancers (Gleason 8). Using a cutoff of 3.0 ng/ml increased these sensitivities to 32 and 68 percent, respectively. The estimated specificity was 91 percent for a PSA cutoff of 4.0 ng/ml and 85 percent for a 3.0 ng/ml cutoff. PSA has poorer discriminating ability in men with symptomatic benign prostatic hyperplasia [45]. Positive predictive value The test performance statistic that has been best characterized by screening studies is the positive predictive value: the proportion of men with an elevated PSA who have prostate cancer. Overall, the positive predictive value for a PSA level >4.0 ng/ml is approximately 30 percent, meaning that slightly less than one in three men with an elevated PSA will have prostate cancer detected on biopsy [40,46,47]. For PSA levels between 4.0 to 10.0 ng/ml, the positive predictive value is about 25 percent [46]; this increases to 42 to 64 percent for PSA levels >10 ng/ml [46,48]. However, nearly 75 percent of cancers detected within the "gray zone" of PSA values between 4.0 to 10.0 ng/ml are organ confined and potentially curable [46]. The proportion of organ-confined cancers drops to less than 50 percent for PSA values above 10.0 ng/ml [46]. Thus, detecting the curable cancers in men with PSA levels less than 10.0 ng/ml presents a diagnostic challenge because the high false-positive rate leads to many unnecessary biopsies. Negative predictive value The Prostate Cancer Prevention Trial, which biopsied men with normal PSA levels, estimated a negative predictive value of 85 percent for a PSA value 4.0 ng/ml [49]. Effect of lowering PSA cutoffs Some investigators have suggested using a lower PSA cutoff because some men with PSA levels below 4 ng/ml and normal digital rectal examinations are found to have prostate cancer [50-53]. In a subset analysis from the placebo arm of the Prostate Cancer Prevention Trial, 449 of 2950 men (15.2 percent) ages 62 to 91 years who had consistently normal PSA levels and digital rectal examinations during the seven years of annual screening had prostate cancer on an end-of-study biopsy; 67 (2.3 percent) had high-grade prostate cancer with a Gleason score of 7 or higher [49]. Among men with a PSA concentration between 2.1 and 4.0 ng/ml, 24.7 percent had prostate cancer, and 5.2 percent had prostate cancer with a Gleason score of 7 or higher. These observations indicate that there is not a clear cutpoint between "normal" and "abnormal" PSA levels. The Prostate Cancer Prevention Trial found that for biopsies performed during follow-up in the control group even a PSA cutoff of 1.1 ng/ml would miss 17 percent of cancers, including 5 percent of poorly differentiated cancers [54]. Thus, any choice of PSA cutoff involves a tradeoff between sensitivity and specificity. While lowering the PSA cutoff would improve test sensitivity, a lower PSA cutoff would also reduce specificity, leading to far more falsepositive tests and unnecessary biopsies. It has been projected that if the PSA threshold were to be lowered to 2.5 ng/ml, the number of men defined as abnormal would double, to up to six million in the US [55]. Additionally, many

4 4 de :24 of the cancers detected at these lower levels may never have become clinically evident, thereby leading to overdiagnosis and overtreatment [56]. There is also evidence that diagnosing prostate cancer at low PSA levels does not affect outcome. A study of 875 men undergoing radical prostatectomy found only a limited association between preoperative PSA levels of 2 to 9 ng/ml and cure rates [57]. The disease-free survival curves did not significantly diverge until the preoperative PSA levels reached 7 ng/ml, suggesting that diagnosing cancers at a lower PSA level may be unnecessary. Most of the PSA elevation below 7 ng/ml was attributed to benign hyperplastic tissue. The investigators emphasized the need for a better serum marker to identify early-stage aggressive cancers. Serial PSA measurements Both detection rates and positive predictive values decline substantially with serial testing [58-61]: During four rounds of annual PSA screening in the PLCO trial, the number of cancers detected per 1000 men decreased from 14.2 to 9.3 [60]. Similarly, the positive predictive value of a PSA level >4.0 ng/ml decreased from 44.5 to 34.9 percent. The cancer detection rate for PSA in the ERSPC, which used a 4-year screening interval, decreased from 5.1 percent in the first round of screening to 4.4 percent in the second round [61]. The positive predictive value for a PSA level of 3.0 ng/ml or greater decreased from 29.2 to 19.9 percent. Studies also found that repeated testing increases the likelihood that detected tumors will be clinically organconfined and be moderately or well differentiated [41,60-62] (see 'Frequency and method of screening' below): In the PLCO, the proportion of screening-detected cancers diagnosed at clinical stage I or II increased from 94.2 percent in round one to 98.5 percent in round 2, while the proportion with Gleason scores 7 decreased from 10.0 to 6.8 percent [60]. In the ERSPC, the proportion of clinical stage I and II cancers increased from 81.5 to 96.3 percent, while the proportion of poorly-differentiated cancers decreased from 8.1 to 3.3 percent [61]. Secular trends in the utility of PSA A United States study that looked at the correlation between PSA level and prostate cancer during five-year intervals at a university hospital found that while serum PSA was correlated with prostate cancer stage, grade, and size in the interval from 1983 to 1988, in the interval from 1998 to 2003 it was correlated only with prostate weight (related to benign prostatic hypertrophy) [63]. The authors concluded that in this era of intense screening for prostate cancer, PSA has ceased to be a useful marker, and biopsies in men with an elevated PSA level are only picking up the background prevalence of prostate cancer [63,64]. That is, the same rates of prostate cancer could be found in men of the same age without regard to PSA level, and in many cases the detected tumors would never become clinically significant. (See 'Overdiagnosis' below.) The authors point out that a study that performed saturation prostate biopsies in men with negative sextant biopsies also found no significant association between PSA level and prostate cancer [65]. The results of these studies raise further concerns about the utility of PSA as a marker for clinically significant prostate cancer. Improving the accuracy of PSA Numerous strategies have been proposed to improve the diagnostic performance of PSA when levels are less than 10.0 ng/ml. These strategies include measuring PSA velocity (change in PSA over time), PSA density (PSA per unit volume of prostate), free PSA, complexed PSA, and using age- and race-specific reference ranges [66]. We suggest not routinely using any of these strategies in deciding which men to refer for biopsy. (See "Measurement of prostate specific antigen".) PSA velocity PSA increases more rapidly in men with prostate cancer than in healthy men. The Baltimore Longitudinal Study of Aging (BLSA) found that men with a PSA rate of change (PSA velocity) greater than 0.75 ng/ml/year were at increased risk of being diagnosed with prostate cancer and that PSA velocity was more specific than a 4.0 ng/ml PSA cutoff (90 versus 60 percent specificity) [67]. The study results, though, were based on analyzing the banked serum of only 18 cancer cases. Furthermore, there are significant short-term physiologic variations in the PSA level [68]. Accurately measuring PSA velocity requires three serial readings, ideally with the

5 5 de :24 same assay, obtained over at least a 12- to 24-month period [66,69,70]. However, analyses of more recent clinical data from randomized trials suggest that PSA velocity adds little predictive information to the total PSA: The European Randomized Study of Screening for Prostate Cancer (ERSPC) data from Rotterdam found that PSA velocity was significantly higher in men with prostate cancer than in men with a negative biopsy (0.62 versus 0.46 ng/ml/year) [71]. However, PSA velocity did not independently predict cancer after adjusting for PSA level. Another analysis of pooled ERSPC data from the Netherlands and Sweden similarly found that PSA velocity only slightly improved the predictive accuracy of total PSA (area under the ROC curve 0.57 versus 0.53), but not for detecting high-grade disease [72]. Among the 774 Rotterdam men with a PSA level below 4.0 ng/ml who underwent their first biopsies in the second round of ERSPC, 149 were found to have cancer [73]. PSA velocity did not discriminate between men with cancer and those with negative biopsies. The sensitivity of a PSA velocity cutoff of 0.3 ng/ml/year was only 39 percent, with a false positive rate of 33 percent. In two separate analyses, the Prostate Cancer Prevention Trial reported on the 5519 men from the placebo group who underwent prostate biopsy following at least two PSA measurements in the preceding two or three years [74,75]. While PSA level was a significant predictor of prostate cancer diagnosis on multivariate modeling, incorporating PSA velocity did not add clinically important predictive information to PSA level alone, particularly for PSA values 4.0 ng/ml. When total PSA was less than 4.0 ng/ml and the DRE was normal, a PSA velocity above 0.35 ng/ml/year predicted cancer. However, using velocity would substantially increase the number of unnecessary biopsies while missing more high-grade cancers than would be identified by just lowering the PSA cutoff. A systematic review of PSA velocity, including 12 comparisons with total PSA for predicting prostate cancer diagnosis, found numerous methodological limitations and essentially no evidence supporting the use of PSA velocity for clinical decision-making [76]. Some investigators have argued that PSA doubling time or percent change is a more appropriate measure of PSA kinetics [77]. PSA velocity is correlated with the total PSA level, which increases exponentially before clinical diagnosis. Even though PSA velocity may be independently correlated with cancer diagnosis, it adds little to the diagnostic accuracy of PSA alone [78]. PSA density The PSA density measurement is based upon the observation that prostate cancers can produce approximately 10 times more PSA per volume of prostate tissue than benign conditions [19,79]. PSA density measurements, which adjust PSA values for prostate volume, have been reported to better discriminate between cancer and noncancer groups than PSA levels alone [80]. However, PSA density measurements require transrectal ultrasound or magnetic resonance imaging to assess prostate volume, which limits applicability in primary care settings. Additionally, precisely estimating prostate volume is difficult [69]. Data from a large multicenter screening trial suggested that using a cutoff PSA density of 0.15 ng/ml/cm3 (a commonly-recommended cutoff value) would miss nearly 50 percent of cancers detected in men with a normal digital rectal examination and PSA levels between 4.0 to 10.0 ng/ml [81]. Adjusting the PSA density cutoff value for total PSA level might improve the test sensitivity [82]. Measuring the PSA density of the prostatic transition zone has also been proposed to improve the specificity of PSA since the hyperplastic tissue that can elevate PSA is almost completely localized to this area of the prostate [66]. Using a PSA transition zone density greater than 0.22 ng/ml/cm3 as a biopsy criterion was estimated to reduce the number of negative biopsies by 24.4 percent based upon data from an Austrian screening study [83]. However, given the logistic difficulties of performing density measurements as well as their lack of reproducibility,

6 6 de :24 the transition zone density is not currently accepted for routine clinical practice [66]. Free PSA The observation that PSA exists in a free form as well as bound to macromolecules has been used to develop additional assays to improve test specificity. The ratio of free-to-total PSA is reduced in men with prostate cancer. Investigators have proposed that biopsies be performed only in men with lower ratios. A large multicenter, prospective trial evaluated men 50 to 75 years with PSA levels between 4.0 and 10.0 ng/ml, including 379 with prostate cancer and 394 with benign prostate disease [84]. The cancer detection rate for this PSA range in screening populations is about 25 percent [46]. However, the detection rate increased to 56 percent for men with a free-to-total PSA ratio less than 10 percent [84]. The investigators selected an optimal cutoff of 25 percent as a criterion for biopsy, which would have reduced the number of unnecessary biopsies by 20 percent in their study cohort. However, men with a normal free-to-total PSA ratio still had an 8 percent probability of having cancer, which may not be low enough to convince patients and clinicians to forego biopsy. A meta-analysis came to similar conclusions that free-to-total PSA ratio is generally only clinically helpful at extreme values of the ratio [85]. A separate meta-analysis of free PSA noted considerable variability in free PSA assays, specimen handling, cutoffs, and patient populations [86]. The authors concluded that more research was necessary to determine the optimal cutoff and to accurately assess the diagnostic performance and utility of the test in screening populations. Complexed PSA Another strategy to improve PSA specificity has been to measure complexed PSA (cpsa). Most circulating PSA is bound to alpha-1-antichymotripsin. A study using archival serum found that, at 95 percent sensitivity, cpsa had a specificity of 26.7 percent compared with 15.6 percent for the free-to-total PSA ratio and 21.8 percent for total PSA [87]. A prospective study in 831 men undergoing prostate biopsy found that cpsa was more specific than total PSA [88]. For men with a total PSA concentration between 4 to 10 ng/ml, when a cpsa cutpoint was chosen to achieve a sensitivity of 90 percent, cpsa had a higher specificity than total PSA (13.3 versus 8.6 percent), but it was less specific than percent free PSA and percent complexed PSA (21.5 and 21.9 percent, respectively). For men with a total PSA concentration between 2 to 6 ng/ml, cpsa was more specific than other methods. The marginal benefit of measuring complexed PSA over total PSA remains uncertain. [-2]ProPSA [-2]ProPSA (also known as p2psa) is a specific isoform of the PSA proenzyme propsa. It has been used to increase the detection of prostate cancer for men with PSA values between 2.0 to 10.0 ng/ml. One prospective observational study estimated that using the p2psa assay (which is not available in the United States) could reduce the number of unnecessary biopsies by 7.6 percent while maintaining a sensitivity of 95 percent for detecting prostate cancer [89]. The cohort included 892 men with normal digital rectal examinations, some of whom previously had negative prostate biopsies. Another prospective study of 268 subjects, using the ratio of p2psa over free PSA, estimated about a 35 percent reduction in the number of unnecessary biopsies while maintaining 95 percent sensitivity [90]. Neither study presented data on the performance of p2psa for detecting high-risk cancers. The clinical utility of this biomarker is uncertain [91]. Age-specific reference ranges PSA levels increase with age, largely due to a higher prevalence of benign prostatic hyperplasia [92]. Although we do not recommend their use, age-specific reference ranges have been developed from normal populations to improve the discriminating power of PSA [93]: 40 to 49 years 0 to 2.5 ng/ml 50 to 59 years 0 to 3.5 ng/ml 60 to 69 years 0 to 4.5 ng/ml 70 to 79 years 0 to 6.5 ng/ml Raising the PSA biopsy threshold in older men improves specificity, reducing the number of unnecessary biopsies. Conversely, lowering the threshold in younger men improves sensitivity and increases detection of early-stage tumors. A retrospective analysis of a large screening cohort reported that applying age-specific reference standards would miss 47 percent of clinically localized cancers in men 70 and older and lead to a 45 percent increase in unnecessary biopsies for men in their 50s [94]. The clinical utility of age-specific reference ranges remains uncertain, and they are not recommended by the US

7 7 de :24 Food and Drug Administration (FDA) or PSA assay manufacturers [66]. Race-specific reference ranges Black men in the United States have the world's highest incidence of prostate cancer and are the most likely to present with advanced stage disease [11]. PSA levels in blacks are higher compared with whites even after adjusting for age, clinical stage, and histology [95]. This difference has been attributed to blacks having larger tumor volumes across all clinical stages. Although we do not recommend their use, race-specific PSA reference ranges have been established in the hope of achieving earlier diagnosis [96]: 40 to 49 years 0 to 2.5 ng/ml (whites); 0 to 2.0 ng/ml (blacks) 50 to 59 years 0 to 3.5 ng/ml (whites); 0 to 4.0 ng/ml (blacks) 60 to 69 years 0 to 3.5 ng/ml (whites); 0 to 4.5 ng/ml (blacks) 70 to 79 years 0 to 3.5 ng/ml (whites); 0 to 5.5 ng/ml (blacks) However, a study of 651 men undergoing radical prostatectomy found that the race-specific reference ranges, which raise the cutoff for blacks 50 years and older compared with whites, would be associated with similar or worse outcomes [97]. The clinical utility of the race-specific reference ranges, which have also been developed for Asians [98], remains uncertain. Summary There is no consensus on using any of the PSA modifications, and none of them has been shown to reduce the number of unnecessary biopsies or improve clinical outcomes. The total PSA cutoff of 4.0 ng/ml has been the most accepted standard because it balances the tradeoff between missing important cancers at a curable stage and avoiding both detection of clinically insignificant disease and subjecting men to unnecessary prostate biopsies [39,56,66]. Ongoing efforts are targeted at identifying new serum markers that will have greater diagnostic accuracy for prostate cancer, particularly for aggressive tumors [66,99]. (See "Measurement of prostate specific antigen".) DIGITAL RECTAL EXAMINATION Digital rectal examination (DRE) has long been used to diagnose prostate cancer. Abnormal prostate findings include nodules, asymmetry, or induration. DRE can detect tumors in the posterior and lateral aspects of the prostate gland; an inherent limitation to the digital examination is that only 85 percent of cancers arise peripherally where they can be detected with a finger examination [100]. Stage T1 cancers are nonpalpable by definition. No controlled studies have shown a reduction in the morbidity or mortality of prostate cancer when detected by DRE at any age [101]. The majority of cancers detected by digital examination alone are clinically or pathologically advanced [102]. Thus, the greatest value of DRE may be its use in combination with PSA testing. (See 'Combining PSA and DRE' below.) Test performance Urologists have been found to have relatively low interrater agreement for detecting prostate abnormalities [103]. No data are available for the test performance characteristics of DRE in primary care. Approximately 2 to 3 percent of men 50 or more years old who undergo a single DRE have induration, marked asymmetry, or nodularity of the prostate. In one analysis, an abnormal screening DRE doubled the odds of detecting a clinically important cancer (defined as a having a tumor volume greater than 0.5 ml) that was confined to the prostate [48]. Although screening DRE increased the likelihood of finding early disease, it also increased the odds three- to ninefold of finding extraprostatic extension of tumor (presumably not amenable to curative therapy). Sensitivity and specificity A meta-analysis of DRE estimated a sensitivity for detecting prostate cancer of 59 percent and a specificity of 94 percent [104]. Positive predictive value The positive predictive value of an abnormal DRE for prostate cancer varies from 5 to 30 percent [46,102, ]. A meta-analysis calculated an overall positive predictive value of 28 percent [104]. COMBINING PSA AND DRE Prostate specific antigen (PSA) and digital rectal examination (DRE) are somewhat complementary, and their combined use can increase the overall rate of cancer detection [39,46, ]. As an example, a multicenter screening study of 6630 men reported a detection rate of 3.2 percent for DRE, 4.6 percent for PSA, and 5.8 percent for the two methods combined [46,106]. PSA detected significantly more of the cancers than digital examination (82 versus 55 percent). Overall, 45 percent of the

8 8 de :24 cancers were detected only by PSA, while just 18 percent were detected solely by digital examination. Investigators reported a positive predictive value of 10 percent for a suspicious digital examination when the PSA level was normal. However, the positive predictive value was 24 percent for an elevated PSA level with a normal digital examination. Among men with a normal PSA level, abnormalities on DRE appear less likely to be from a cancer if the PSA concentration is below 1.0 ng/ml than if the PSA concentration is between 3.0 to 4.0 ng/ml [108]. Although these data suggest a potential benefit for combining PSA and DRE in detecting prostate cancer, randomized trials have not confirmed a benefit on prostate cancer outcomes. The ERSPC, which found a small survival benefit with PSA screening, did not consistently require DRE [13]. The PLCO found no survival benefit with combined PSA and DRE screening [15]. OTHER TESTS PCA3 The prostate cancer antigen 3 gene (PCA3), which was identified in 1999, is highly overexpressed in almost all prostate cancer tissue specimens, but not in normal or hypertrophied tissue [112]. A PCA3 score, based on the ratio of PCA3 mrna over PSA mrna (which is not related to serum PSA levels or cancer), can be determined from a urine specimen collected after a vigorous digital rectal examination. PCA3 has been evaluated for guiding biopsy decisions when PSA levels are in an indeterminate range (2.5 to 10.0 ng/ml) and for men with previously negative biopsies but persistently elevated PSA levels. A 2010 review identified 11 clinical trials, representing 2737 subjects, evaluating the diagnostic performance of PCA3 [113]: In four studies evaluating patients with indeterminate PSA, sensitivity ranged from 53 to 84 percent and specificity ranged from 71 to 80 percent. In three studies with at least 200 patients that provided data on PCA3 performance following a previous negative biopsy, sensitivity ranged from 47 to 58 percent, and specificity ranged from 71 to 72 percent. PCA3 outperformed PSA and percent free PSA in independently predicting a positive biopsy. However, determining the clinical utility of PCA3 from these studies is difficult. Aside from the relatively small sample sizes, studies differed in their criteria for biopsy referral (PSA levels 2.5 to 3.0 ng/ml, digital rectal examination findings, or risk factors), the generation of the PCA3 test used, and the cutpoint for defining an abnormal test. Additionally, none of the studies used PCA3 scores as an indication for biopsy. The Rotterdam site of the ERSPC subsequently reported the results of using PCA3 as an initial screening test, with sextant biopsy performed if either the PSA level was 3 or the PCA3 score was 10 [114]. Based on receiver operating characteristic (ROC) curve analysis of 721 subjects undergoing biopsy, PCA3 performed only marginally better than total PSA (area under the curve 0.64 versus 0.58, p = 0.14); PCA3 also missed the majority of cancers with Gleason >6 or stage T2a, though only 19 men met these criteria. However, the generalizability of these results is uncertain because all subjects had already undergone three rounds of screening, and 29 percent had previous negative biopsies. While PCA3 may eventually have a role in reducing unnecessary biopsies, there are insufficient data on clinical outcomes to currently support routine use. Transrectal ultrasonography Transrectal ultrasonography (TRUS) is an outpatient procedure that requires no sedation or analgesia and is relatively well tolerated by most men. TRUS is not recommended as a primary screening test for prostate cancer because of its low sensitivity and positive predictive value. As an example, in one study almost 40 percent of cancers would have been missed if prostate biopsies had been performed only in men with suspicious findings on TRUS [46]. Furthermore, TRUS is not a feasible screening test in primary care clinics. TRUS is typically used to guide prostate biopsy rather than as a screening test. EFFECTIVENESS OF PROSTATE CANCER SCREENING Apart from issues of cost and acceptability, in order for prostate cancer screening to be valuable, it must reduce disease-specific morbidity and/or mortality. Evidence from randomized trials Two well-designed large randomized trials have evaluated the effectiveness

9 9 de :24 of screening for prostate cancer and found somewhat differing results: In the European Randomized Study of Screening for Prostate Cancer (ERSPC), 182,160 men between the ages of 50 and 74 were randomly assigned to PSA screening (an average of once every four years) or a control group that was not offered screening [13]. This study used different recruiting and randomization procedures across seven centers in Europe. The study used PSA cutoffs between 2.5 and 4.0 ng/ml (most centers used a cutoff of 3.0 ng/ml) as indications for referral for biopsy, variably supplemented with DRE, transrectal ultrasonography, and/or measurements of free PSA levels. The overall rate of prostate cancer screening in the control group was not reported, though 31 percent of cancers were categorized as stage T1c (diagnosed based on elevated PSA level). Investigators subsequently reported PSA testing among 24 percent of the Rotterdam site controls and estimated that 50 percent of the tests were for screening [115]. After a median follow-up of 11 years, for the 162,243 men between the ages of 55 and 69 the primary outcome of prostate cancer mortality was 21 percent lower in the group offered screening (rate ratio 0.79, 95% CI ) [116]. The absolute rates of prostate cancer mortality were 0.39 versus 0.50 per 1000 person-years (absolute reduction of 0.10 deaths per 1000 person-years; 1055 men needed to be invited for screening to prevent one prostate cancer death over 11 years). Prostate cancer was diagnosed more frequently in the screening group (9.7 versus 6.0 cases per 1000 person-years), such that 37 additional cases of prostate cancer would need to be detected by screening to prevent one death from prostate cancer over 11 years. All-cause mortality was not reduced with screening (18.2 versus 18.5 deaths per 1000 person-years; rate ratio 0.99, CI ). Prostate cancer mortality was also reduced in the entire cohort of men ages 50 to 74 (rate ratio 0.83, CI ). Several centers collected quality-of-life data; however, these results have not yet been published. Although the absolute mortality benefit for screening was low, several factors could have biased the results toward no effect. About 24 percent of subjects invited for screening did not undergo PSA testing [13]. While not definitively characterized, a substantial proportion of the control group likely received PSA testing (31 percent of cancers were screening-detected). A subsequent analysis of the Rotterdam site data used patient surveys and linkages with a central national laboratory to estimate contamination. Adjusting for contamination and non-adherence with screening, investigators estimated that prostate cancer screening could reduce prostate cancer mortality by as much as 31 percent (95% CI 8-49 percent) [117]. Additionally, at least 25 percent of cancers detected in the screening group did not receive curative treatment with either surgery or radiation. Finally, given the indolent course of prostate cancer and the five to ten year lead time associated with PSA testing, follow-up duration may have been insufficient to accurately estimate the survival benefit. A modeling study using data from all ERSPC sites concluded that the screening benefit could increase over time, with numbers needed to screen of 837 at year 10 and 503 at year 12 [118]. However, any survival benefit from screening would not be realized for many years, while the burdens of screening and treatment, including harms from overdiagnosis and overtreatment, would occur immediately and potentially have lifelong consequences. Several biases could also have favored the screening group [119]. A higher proportion of cancers diagnosed in the screening group were aggressively treated (surgery or radiation) compared to the control group, so some of the outcome differences could be related more to improved treatment than screening. Additionally, the committee adjudicating cause of death was aware of cancer treatments. Previous studies have suggested that cause of death is less likely to be attributed to prostate cancer for patients who received aggressive treatment [120]. The ERSPC investigators did not report the association of cancer death and receipt of treatment. The ERSPC site from Göteborg, Sweden subsequently reported a cumulative mortality reduction of 44 percent for the PSA screening group after a median 14 years of follow up [121]. The Göteborg site, which used population registries to randomly allocate men to either the screening or control groups, could plausibly be more likely to demonstrate benefit than the other ERSCP sites because it offered screening every two years (versus every four years) and the median follow up was 5 years longer than for the initial report. Additionally, the Göteborg results were based on a cohort of men ages 50 to 64, compared to ages 55 to

10 10 de :24 69 in the combined ERSPC report. This suggests that screening may be less beneficial for men 65 and older, consistent with the finding that radical prostatectomy did not confer a survival benefit compared to watchful waiting for men in this age range [122]. Nevertheless, the 95 percent confidence interval for mortality reduction ranged from 18 to 61 percent, which overlaps the overall 20 percent risk reduction previously reported by ERSPC, and the absolute cumulative risk reduction was only 0.40 percent (95% CI %). Screening also increased the risk for prostate cancer diagnosis by 55 percent. Finally, men in the screening group were more likely to receive attempted curative therapy than those in the control group, particularly radical prostatectomy. In the United States Prostate, Lung, Colorectal and Ovarian Cancer (PLCO) Screening Trial, 76,693 men between the ages of 55 and 74 were randomly assigned to annual screening with PSA and DRE or to usual care [15]. A PSA level above 4.0 ng/ml or an abnormal DRE were indications for biopsy. A high proportion of men in the control group underwent PSA testing (52 percent in the sixth year of the study) and over 40 percent of study subjects had undergone PSA testing within three years before enrolling in the trial. In contrast to the ERSPC, after seven years of follow-up there was no reduction in the primary outcome of prostate cancer mortality (50 versus 44 deaths in the screening and control groups, respectively; rate ratio 1.13, 95% CI ). Cancer detection in the screening group was significantly higher than in the control group (2820 versus 2322, rate ratio 1.22, CI ). A subsequent publication looking at longer-term follow-up within the PLCO (92 percent follow-up through 10 years; 57 percent through 13 years) found similar prostate cancer mortality results (RR 1.12, CI ) with no suggestion of reduced mortality in the patients followed for 13 years (RR 1.09, CI ) [123]. This suggests that the differences in results between the ERSPC and the PLCO were not related to the duration of follow-up. Additionally, the investigators found no evidence that screening could be beneficial in any subgroups defined by comorbidity, age, or pretrial PSA testing. The negative results could be attributable to the high rate of PSA testing in the control arm, the high proportion of subjects with recent PSA testing at baseline (because serial testing is associated with finding fewer and less aggressive cancers), the higher PSA cutoff for biopsy compared with that used in the ERSPC, or the small number of prostate cancer deaths. An earlier PLCO publication also indicated that substantial proportions of men with abnormal PSA and/or DRE results had not undergone biopsy within three years following the positive screen [60]. All of these factors could bias the PLCO trial toward a null result, and also suggest that further follow-up is not likely to yield positive results. One earlier randomized trial of screening for prostate cancer reported positive findings, but the data analysis was flawed. In this population-based study in Quebec, 46,193 men aged 45 to 80 years identified from electoral records were randomly assigned to screening with prostate specific antigen (PSA) and digital rectal examination (DRE) versus no screening [58]. In an analysis that excluded the 77 percent of men in the screening arm who declined screening and excluded the 6.5 percent of men in the control group who were screened, the prostate cancer mortality rate in men undergoing screening was reported to be 67.1 percent lower than in the control group. When the data were evaluated by a more appropriate intention-to-screen analysis, there were no mortality differences between the two groups (4.6 versus 4.8 deaths per 1000 persons, respectively). Additionally, the results suggesting benefit seemed biologically implausible, since the survival benefit became apparent within only three years, a very short time for a screening program to be effective given the long lead time for prostate cancer. A 2010 meta-analysis summarized results from six randomized trials (including unique data from two ERSPC sites), with a total of 387,286 participants [124]. Screening with PSA with or without DRE compared to no screening did not reduce death from prostate cancer (relative risk [RR] 0.88, 95% CI ). However, screening significantly increased the probability of cancer diagnosis (RR 1.46, CI ). In a 2011 Cochrane meta-analysis that had similar findings, the estimated prostate cancer-specific mortality difference was not statistically significant (RR 0.95, 95% CI ), but cancer was diagnosed significantly more often in men randomized to screening (RR 1.35, 95% CI ) [125]. The concerns that ERSPC mortality findings favoring screening could have been due to differential treatment

11 11 de :24 between the screening and control arms [119,126] have been partially addressed by results from the Prostate Cancer Intervention versus Observation Trial (PIVOT) [127]. PIVOT randomly assigned men with localized prostate cancer, the majority of whom had been detected by PSA screening, to either radical prostatectomy or watchful waiting. After a median follow-up of 10 years, men who were assigned to radical prostatectomy had a reduction in prostate cancer mortality that was not statistically significant (5.8 versus 8.4 percent; HR 0.63, 95% CI ). Subgroup analyses suggested a potentially greater survival benefit for radical prostatectomy among men with PSA values above 10 ng/ml or high-risk tumor characteristics (based on tumor stage, Gleason, PSA) compared with the group as a whole. (See "Radical prostatectomy for localized prostate cancer", section on 'Survival impact of radical prostatectomy'.) Evidence from observational studies Before publication of the randomized trials, other data had been cited to support the effectiveness of screening. Given the conflicting results discussed above, observational studies provide information that can fill in some gaps in evidence from the trials. Over the past decade, Surveillance Epidemiology and End Results (SEER) tumor registry data have shown a significant decline in the incidence of advanced stage disease, potentially consistent with effective screening [128]. Prostate cancer mortality rates, which initially increased following the advent of PSA testing, have now declined to slightly below pre-psa levels (figure 1) [128]. These mortality trends, however, are difficult to interpret. Some ecologic data suggest an association between PSA testing and declining mortality rates: In Austria, the state of Tyrol introduced mass screening in Within five years, investigators observed a more than threefold adjusted decrease in prostate cancer mortality rates in Tyrol compared with the rest of the country where screening was less common [129]. Data from Olmsted County in Minnesota demonstrated that age-adjusted 1993 to 1997 prostate cancer mortality rates declined 22 percent (95% CI 49 to 17) compared with rates measured in years before PSA testing [130]. A study found that age-adjusted prostate cancer mortality peaked in the early 1990s in both the United States and the United Kingdom, but then declined much faster in the US, where PSA screening became common, than in the UK, where PSA screening was less common (yearly decline -4.2 versus -1.1 percent) [131]. However, other ecologic studies have shown declining mortality rates even in the absence of intensive screening: Trends in prostate cancer mortality rates in Wales and England from 1991 to 1997 were comparable to trends in the United States. Mortality rates declined by 1.7 percent in the United Kingdom, even though PSA screening was not routinely performed and even discouraged [132]. Regional practice variation has allowed investigators to evaluate the effect of screening and treatment on prostate cancer mortality within the United States: Medicare beneficiaries in Seattle received more intensive PSA screening and aggressive cancer treatment from 1987 to 1990 than beneficiaries in Connecticut. However, there were no differences in prostate cancerspecific mortality over 11 years of follow-up; the adjusted mortality rate ratio was 1.03 (95% CI ) in Seattle compared with Connecticut [133]. Alternative explanations have been proposed for declining mortality rates. Better primary treatments could reduce mortality rates among men diagnosed with localized cancer. Additionally, the use of androgen deprivation therapy and other chemotherapies for men with advanced-stage cancer could allow men to survive long enough to die from a comorbid condition. (See "Initial management of regionally localized intermediate and high risk prostate cancer" and "Overview of the treatment of disseminated prostate cancer".) Evidence from modeling studies Simulation models using data from Surveillance Epidemiology and End Results (SEER) registries suggest that PSA screening could account for 45 to 70 percent of the observed decline in prostate cancer mortality rates, mainly by decreasing the incidence of distant stage disease [134]. However,

12 12 de :24 treatment advancements may have also contributed to the declining mortality rates. The European Randomized Study of Screening for Prostate Cancer (ERSPC) investigators used simulation models based on their data and observational studies reporting quality of life outcomes to project lifetime numbers of cancer diagnoses, treatments, deaths, and quality-adjusted life years gained after PSA screening [135]. Overall, annual screening between ages 55 to 69 would result in nine fewer prostate cancer deaths per 1000 men followed for an entire lifetime, with a total of 73 life-years gained. Investigators projected that 98 men would need to be screened and five cancers detected to prevent one prostate cancer death. However, after adjusting for the adverse effects of screening, PSA screening resulted in an estimated loss of 16.7 quality-adjusted life-years, ranging from 24.4 life-years gained to 93.8 life-years lost. An editorialist noted that these results demonstrate that screening decisions are very sensitive to patient preferences for potential future health states [136]. A study used microsimulation modeling of observational and clinical trial data to try to determine the comparative effectiveness of alternative PSA screening strategies [137]. Outcome measures included the lifetime number of PSA tests, false-positive results, cancer detection, overdiagnosis, prostate cancer deaths, and lives saved. Compared to a reference strategy of annual PSA testing between ages 50 to 74 with a PSA threshold of 4.0 ng/ml for biopsy referral, strategies that stopped screening at an earlier age, widened testing intervals, and/or used age-adjusted PSA biopsy criteria appeared to reduce the number of tests and the risks for false-positive results and overdiagnosis, while increasing the absolute risk of prostate cancer death by only a fraction of one percentage point. Conversely, screening strategies that lowered the starting age and/or PSA threshold for biopsy referral appeared to markedly increase the number of tests and the risks for false-positive results and overdiagnosis, while only marginally decreasing the risk of prostate cancer death. However, concerns were raised about the analyses, including the failure to model risk factors, the use of simplified measures for stage and grade, and not considering patient preferences [138]. HARM FROM SCREENING Risks of biopsy Although early reports indicated that prostate biopsies very rarely (<1 percent) caused complications (eg, bleeding, infection) serious enough to require hospitalization [139], more recent studies suggest both higher rates of infectious complications and that the rate of infectious complications may be increasing over time [ ]. Hospitalization rates for infectious complications in these studies have ranged from 0.6 to 4.1 percent [142]. Infectious complications can lead to sepsis, which can very rarely lead to death. A modeling study, assuming a biopsy mortality rate of 0.2 percent [143], concluded that prostate cancer screening could be associated with a net increased overall mortality, particularly under the conditions that biopsy rates are high and screening is relatively ineffective [144]. However, other studies have suggested much lower mortality rates following biopsy [142]. Population-based studies include an analysis of US Medicare data that found a mortality rate of 0.3 percent in the 30 days following biopsy; this was actually 70 percent lower than the 30-day mortality in a comparison population not undergoing biopsy [140]. An analysis of registry data from Canada found a 30-day mortality rate of 0.09 percent [141]. Randomized trials with follow-up on 1147 biopsies [145], and 10,474 biopsies [146], reported no biopsy-related deaths. Prostate biopsy can also lead to anxiety and physical discomfort [147]. Among 116 men undergoing biopsy in the Rotterdam screening study, 55 percent reported discomfort with the procedure, including 2 percent who had pain persisting longer than one week. Being diagnosed with prostate cancer is psychologically distressing, but even patients with a negative biopsy result may be distressed [148,149]. Chronic anxiety can follow a negative prostate biopsy because this apparently favorable result cannot completely rule out prostate cancer given the relatively high false-negative biopsy rate [150]. Overdiagnosis Overdiagnosis refers to the detection by screening of conditions that would not have become clinically significant. When screening finds cancer that would never have become clinically significant, patients are subject to the risks of screening, confirmatory diagnosis, and treatment, as well as suffering potential psychosocial harm from anxiety and labeling. Overdiagnosis is of particular concern because most men with screening-detected prostate cancers have early-stage disease and will be offered aggressive treatment.