We asked these authors, generally perceived as skeptical about prostate cancer screening, what they think current policy for prostate cancer screening should be. We then asked some follow-up questions.
Peter C Albertsen (Fig. 5) and Dragan Ilic (Fig. 6)
A Vickers and S Carlsson: You have generally been perceived as an individual who is skeptical about prostate cancer screening. What do you think current policy should be; should there be no screening ever? That is, should PSA be avoided entirely in asymptomatic men, or should screening be restricted to certain subgroups and, if so, whom? Justify your answer with reference to the literature.
P Albertsen: Few cancers generate as much controversy as prostate cancer concerning screening, diagnosis, and treatment. From 1977 to 2005, the lifetime risk of prostate cancer diagnosis in the US increased from 7.3 % to 17 % [31, 32]. During this same period, the lifetime risk of dying from prostate cancer fell from 3.0 % to 2.4 %.
My views on prostate cancer screening and treatment have been shaped by my training in urology at Johns Hopkins and epidemiology and public health at the University of Wisconsin, as well as by my patients. One patient in particular had a powerful influence. He was referred for treatment of a localized prostate cancer and underwent successful surgery. His postoperative PSA was undetectable and all surgical margins were negative. I told him he was cured only to have him return 2 years later with widespread metastases. I treated him successfully with anti-androgen therapy for an additional 16 years. According to the Halsted model of cancer progression he should have been cured by surgery [33]. I had the original specimen re-cut to document negative margins and concluded that we had a poor understanding of the natural history of this disease and the efficacy of treatment.
To address this problem, I developed a Markov model of prostate cancer progression. With the assistance of Jack Wennberg’s research group, the model was published in JAMA [34]. At the time, it was severely criticized by the urologic community, but when reviewed today the model bears a remarkable resemblance to data recently published by the SPCG-4 [35]. The sensitivity analysis indicated that data concerning the natural history of this disease was most critical to estimating the relative value of intervention. This is why I gathered data on the natural history of this disease from patients enrolled in the Connecticut tumor registry and published them in JAMA in 1998 and 2005 [36, 37].
My training in epidemiology taught me to view healthcare delivery from a public health perspective. Screening programs were best assessed by addressing four key questions: (1) Is prostate cancer a suitable disease for screening? (2) Is treatment for prostate cancer effective? (3) Is PSA an effective screening test for this disease? (4) Does screening result in any harm?
Early detection and treatment for prostate cancer is not a new idea. As early as 1905, Hugh Hampton Young suggested that a careful digital rectal examination could identify changes in prostate gland texture that could lead to the early diagnosis of cancer and appropriate intervention [38]. Only recently have we begun to appreciate that a large number of men harbor indolent disease, as clearly demonstrated by data from the Finasteride Chemoprevention trial [39]. Many pathologists now question whether Gleason 3 + 3 tumors are sufficiently aggressive to cause morbidity [40]. Indeed, recent data from the ERSPC trials suggest that half of all screen-detected cancers are indolent [5].
The efficacy of treatment also poses problems. Most urologists and radiation therapists assume that surgery and radiation are curative. However, what do the data say? The SPCG-4 suggests that some men are cured by surgery, but many men are not [35]. Further, men with high-grade disease often die from prostate cancer despite surgery and that surgery is primarily palliative for men aged over 65. With regards to radiation, even less information concerning its efficacy is available to date.
How well does PSA perform as a screening tool? Unfortunately, many lethal tumors produce low amounts of PSA and are missed by screening studies. Additionally, prostate enlargement, prostatitis, and surgical manipulation can lead to a significant number of false positive values. Finally, screening can result in considerable morbidity as documented by the USPSTF report [1].
Do I believe that PSA testing should be abandoned? Of course not, screening clearly benefits some men. This is why I helped draft and support the prostate cancer screening guidelines approved by the AUA [6]. Unfortunately, most clinicians focus primarily on the potential benefits of screening in individual patients and downplay the harms associated with testing as measured from a public health perspective. Before we can advocate for population-based screening, we must understand the natural history of the disease identified by testing and gain a better appreciation of the efficacy of treatment. This is why I have been an advocate of active surveillance for men with low-grade disease [9]. Data from the ProTECT trial should provide important new data concerning efficacy when published in the spring of 2016 [41]. Ideally, screening should only identify men destined to suffer from clinically significant disease and patients should only be offered treatments that yield substantial benefit. We still have a long way to travel to reach this goal.
D Ilic: Prostate cancer is a leading cancer affecting men worldwide [42]. Despite its high prevalence, the current evidence suggests that screening asymptomatic men for prostate cancer is not warranted [1, 43]. The most recent Cochrane systematic review identified five randomized controlled trials examining the effectiveness of screening [43]. A meta-analysis of data from those five trials determined no significant difference in prostate cancer mortality between men randomised to screening in comparison to those who were not (risk ratio (RR) = 1.00; 95 % CI, 0.86–1.17) [43].
Only two of the five randomized controlled trials included in the 2013 systematic review were determined to methodologically present a low-risk of bias [43]. The point of interest lies in the differing results and conclusions offered by two studies: the ERSPC and the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial [27, 44]. The ERSPC study was the only study, out of the five included in the meta-analysis, to demonstrate a reduction in prostate cancer mortality from screening (RR = 0.84; 95 % CI, 0.73–0.95) [27]. Conversely, the PLCO study reported no benefit in screening (RR = 1.15; 95 % CI, 0.86–1.54) [44]. Criticisms of both the ERSPC and PLCO studies exist. The PLCO study has been criticized for the large number of participants entering the trial with a history of screening and the level of contamination by control participants continuing to be screened [43, 45, 46]. The ERSPC study has been critiqued for the variation in study protocols and their implementation across study sites [43, 46]. In 2014, the ERSPC published 13-year follow-up data, reporting a 21 % reduction in the risk of prostate cancer mortality through screening (RR = 0.79; 95 % CI, 0.69–0.91) [5]. A sub-group analysis of prostate cancer mortality by age at randomization identified a significant decrease in prostate cancer mortality in the 65–69 year age group (RR = 0.69; 95 % CI, 0.55–0.87). No statistically significant difference in prostate cancer mortality was observed between screening and control groups in men aged <54, 55–59, and 60–64 years [5]. The results also suggest that screening is not beneficial in men aged over 70 years (RR = 1.17; 95 % CI, 0.82–1.66). The ERSPC study authors concluded that, “…the time for population-based screening has not yet arrived…” [5].
Given that the current evidence does not support population-based prostate cancer screening, the question then turns to screening on an individual basis. For individual patients to make an informed decision, they must be aware of the benefits and harms associated with the diagnostic tests used when screening for prostate cancer. The PSA test is used as the common frontline test for prostate cancer; nevertheless, its sensitivity and specificity, as reported in the literature [47], varies widely. The potential for high false positive and significant false negative rates with PSA testing can lead to substantial harms, including overdiagnosis and overtreatment [1, 43, 47, 48]. Identifying a cancer that would never have become apparent in the absence of screening and then subjecting the patient to invasive treatment can lead to significant physical, emotional, and psychosocial harms [1, 43].
Age, family history, and ethnicity have been determined as risk factors for prostate cancer [49]. Whilst there is no evidence to support population-based screening, patients and clinicians should discuss the potential benefits and harms of screening as it relates to the individual patient [1, 43]. The use of decision aids and risk calculators can increase patients’ knowledge and decision-making to ensure that patients and clinicians engage in an informed, evidence-based decision grounded on the patient’s needs and circumstances [50].
A Vickers and S Carlsson [to P Albertsen]: You argue that, “
before we recommend ‘population-based screening’ we must understand the natural history of the disease identified by testing and gain a better appreciation of the efficacy of treatment.
” Can you be a little bit more specific?
What are the precise gaps in our understanding of the natural history of prostate cancer that prevent us from developing optimal screening programs?
What research, exactly, must be performed to address these gaps? What are the specific gaps and associated research studies that need to be carried out as regards treatment efficacy?
P Albertsen: For a screening program to be successful from a public health perspective (i.e. population-based screening) three critical issues must be satisfied: (1) the disease in question must pose a significant health burden, (2) the screening tool must be able to identify the disease sufficiently early in its natural history so that outcomes can be altered, and (3) a treatment must exist that can alter the outcome of the disease identified by screening. Many urologists believe that PSA testing satisfies these criteria; unfortunately, it does not – neither when it was proposed in the late 1980s, nor now.
PSA screening does increase the likelihood of finding a condition termed prostate cancer, but it does not discriminate between a disease representing a significant health burden and one that is essentially indolent. As noted in the most recent report by the ERSPC, approximately half of the cancers identified by PSA testing are Gleason 6 and are not destined to pose any significant clinical burden [5]. We also know that PSA testing among older men is unlikely to identify clinically significant disease in many men, yet most screening appears to be done in this age group [51]. The Finasteride Chemoprevention trial has provided critical information on the widespread prevalence of low-grade prostate cancer and the PIVOT trial and the SPCG-4 trial have provided much needed information on the natural history of prostate cancer [35, 52, 53] The ProTECT trial should add to this critical data when reports are published next year. While many of us recommend active surveillance for men with Gleason 6 disease, most of these men are still receiving aggressive interventions that often do more harm than good.
Finally, the third criterion is equally important. Most urologists assume that a radical prostatectomy is curative; nevertheless, data to support this assertion are missing. Currently, the best data have arisen from the SPCG-4 trial, which demonstrates that most men with high-grade disease do not benefit from surgery [35]. Prostate cancer mortality is similar for men undergoing surgery and those undergoing delayed androgen suppression. Surgery does appear to benefit men with intermediate-grade disease, but this is seen primarily among younger men rather than those aged over 65. It could be that the study investigators did not identify men sufficiently early during the course of their disease; however, the PIVOT trial has also raised questions regarding the efficacy of surgery [53], thus highlighting the importance of the ProTECT trial. Unfortunately, we have even less information concerning the efficacy of radiation in any of its forms [41].
What are the gaps in our understanding of the natural history of prostate cancer? We need a tool that can clearly indicate whether a man has a disease that will cause significant morbidity and death during his natural lifetime. We also need randomized trials that demonstrate how well our primary treatments, namely surgery and radiation, work to alter the natural course of prostate cancer. There is no substitute for this information.
A Vickers and S Carlsson: One of the problems with waiting for new data is that it is rarely as definitive as anticipated. You say that the “
ProTECT trial should provide important new data
” in 2016, but that will be pretty early data – only 10 years of follow-up – and does not reflect best practice: a single PSA when a man is in his 50s or 60s, surgery from low-volume surgeons, and relatively low-dose radiotherapy. As regards the treatment trial in ProTECT, about three quarters of patients have low-grade disease, which we now believe should not be treated at all. Do you really think that ProTECT will provide enough data about screening and treatment to move forward with practice recommendations?
P Albertsen: We currently have practice recommendations from the AUA that focus PSA testing on the group of men most likely to benefit. While I agree that research trials are rarely as definitive as anticipated, each new study helps refine our understanding of the efficacy of screening and treatment. The ProTECT study will publish 10-year data next year. I agree that this is a relatively short follow-up, but it is 10 years ahead of any randomized trial being conducted in the US. You state that the trial does not reflect best practice, but how can you be sure that PSA testing, as currently practiced in the US, is best practice (see response to question #1). You dismiss the trial because it is being conducted by ‘low-volume surgeons’. What evidence do you have that surgery is less efficacious among low-volume compared to high-volume surgeons? These surgeons know how to operate and have undergone quality assurance comparisons as part of the trial. Is that also true for high-volume surgeons in the US? Imagine that the high-volume surgeons had participated in a similar trial when PSA testing became widely adopted in the US. If this had been the case, we would now have a completed trial with a large sample size and 20-year follow-up. The UK government would rather invest in clinical trials to determine if screening and treatment works, than simply pay for surgery and radiation that urologists/radiation therapists claim has a major impact. Finally, I would be surprised to learn that three-quarters of the men enrolled in the trial have low-volume, low-grade disease. These men are usually identified following repeated PSA testing. Let’s review the results together in 2016. Without the European trials, including SPCG-4, ERSPC, and ProTECT, we would only be left with the PIVOT and PLCO data [54]. How would urologists debate prostate cancer screening and treatment with these data? You seem to advocate crafting screening and treatment policies with the hope that the tools and treatments work. I am a bit more skeptical given the history of breast cancer screening and management and that of other medical theories that have been discarded.
A Vickers and S Carlsson: You say that “
many lethal tumors … are missed by screening
”. We all know that does sometimes occur, but do you have any good estimates as to how common it is? The Malmö studies suggest that the incidence of lethal disease in men with low PSA is actually extremely low.
P Albertsen: I agree with your statement that the incidence of lethal disease in men with low PSA is extremely low. I also believe that the incidence of high-grade, lethal prostate cancer, is also very low in general. This latter statement carries significant implications on the performance of a screening test. When the incidence of the condition you are searching for is low and the pool of potential patients is very large, test performance becomes critical – false positives rapidly overwhelm true positives. You can reassure most men in their early 50s and even 60s that they will not die from prostate cancer in the absence of PSA testing and be right 97 out of 100 times because the lifetime risk of dying from prostate cancer is approximately 3 %. Unfortunately, we have all seen the young men in their 50s who present with Gleason 8–10 tumors and have extra prostatic extension by the time they undergo surgery. As a clinician, I wish we could find these tumors earlier and cure these men. As someone trained in epidemiology, I know that the ability to achieve this dream is very difficult, especially when the tests have modest sensitivity and specificity and the treatments are moderately effective (see responses to Questions 1 and 2) [1].
A Vickers and S Carlsson: The original question asked “
Should PSA be avoided entirely in asymptomatic men, or should screening be restricted to certain subgroups and, if so, whom?
”. You seem to agree with the latter: you say that screening should not be abandoned because it “
clearly benefits some men
”. So which men should be screened?
P Albertsen: Based upon what we know in 2015, PSA testing does benefit some men. For this reason I strongly support the recommendations of the AUA. These guidelines were based extensively on the data provided by the ERSPC. However, we need to do better. I am aware of your work with Hans Lilja, suggesting that a PSA value at age 50 is predictive of the long-term probability of developing clinically significant prostate cancer [18]. I believe we can refine the group of men who should be tested. We also need to incorporate the natural increase in PSA that comes with prostate enlargement that occurs when a man ages through his 50s and 60s. A graphic chart that tracks PSA levels or, possibly, percent free-PSA or the new prostate health index test against age and prostate volume, similar to a pediatric growth chart, might be helpful. Men consistently falling outside the 90th percentile, for example, might undergo MRI testing before considering a prostate biopsy. All of these refinements should be aimed at lowering the incidence of low-volume, low-grade cancer. In my mind, we have yet to define a best practice screening and treatment algorithm. While the AUA recommendations are a start, we have not agreed upon the test(s) needed, the frequency of their application, the value of imaging versus biopsies, nor on which treatments work best for which patients. These are the research gaps that should be addressed by trials. Unfortunately, these trials are often difficult to conduct in the US because of perverse economic incentives and clinicians who are convinced that current practices are the preferred standard of care.
A Vickers and S Carlsson [to D Ilic]:
You seem to base your conclusions on the Cochrane meta-analysis. However, as you mention, three of the studies were associated with a high-risk of bias. Moreover, one of these studies did not even include PSA screening until very late in the trial. Do you think it is valid to report a risk ratio from a combined analysis including poor quality studies?
D Ilic: The meta-analysis of all five studies provides a broad overview of the evidence available on the topic. The Cochrane review cited provides a sensitivity analysis in which these poor quality studies are removed and their impact upon the overall meta-analysis can be identified. It is interesting to note that, even following their removal, the sensitivity analysis with the two ‘good’ quality trials was not greatly affected (RR = 0.96; 95 % CI, 0.70–1.30 (‘good quality trials’) vs. RR = 1.00; 95 % CI, 0.86–1.17 (‘all trials’)) [43].
A Vickers and S Carlsson: You imply that the key consideration is to compare the results of the PLCO and ERSPC studies. These studies addressed very different study questions. The ERSPC was a trial of screening versus no screening; the PLCO was explicitly described by the investigators as a trial of ‘systematic’ vs. ‘opportunistic’ screening. Would we expect these studies to have results that are in any way comparable? Do you think it makes sense to combine the results meta-analytically? By way of analogy, if we had a trial of aspirin vs. no aspirin for pain, and another trial of systematic aspirin vs. aspirin as needed, do you think we could combine the trials to get an estimate for aspirin effectiveness?
D Ilic: The aim of performing a meta-analysis with these two studies is to increase the power, improve precision, and hopefully have the evidence to answer such a controversial question. In performing a meta-analysis, careful consideration is given to the impact of clinical, methodological, and statistical heterogeneity upon the outcome [55]. The PLCO study has been described as outlined above, partly due to the high rate of contamination reported and the potential for such contamination to mask any clinical benefit from screening. Methodological heterogeneity has been reported for the ERSPC study, with significant variation amongst study sites with respect to screening protocol, contamination, and follow-up [5]. In performing a meta-analysis of these studies, the assumption is that their objectives, study design, participants, interventions, outcomes measured, and follow-up are sufficiently homogeneous to allow the pooling of data.
A Vickers and S Carlsson: You say: “
No statistically significant difference in prostate cancer mortality was observed between screening and control groups in men aged less than 54, 55–59, and 60–64 years
”. What conclusion do you draw from that?
D Ilic: The rate ratios (RR) for the respective age groups are as follows, <54 (RR = 0.84; 95 % CI, 0.28–2.49); 55–59 (RR = 0.81; 95 % CI, 0.93–1.03); and 60–64 (RR = 0.90, 95 % CI, 0.71–1.15). These results would indicate that there is no evidence to support the hypothesis that screening significantly reduces the risk of prostate cancer mortality in these groups. For example, the risk ratio and confidence intervals for the <54 age group indicate great variability for the potential of benefit and harm as a result of screening. It should be noted that the AUA recently modified their guideline recommending against PSA screening in men under 40 years of age, and in men aged 40–54 years at average risk. For men aged 55–69 years, a shared approach to decision-making was advocated [56].
A Vickers and S Carlsson: You state that “
patients and clinicians should discuss the potential benefits and harms of screening as it relates to the individual patient
” and then cite two papers: your own review showing no benefit to screening and the USPSTF report that explicitly stated that there was a reasonable degree of ‘certainty’ that screening does more harm than good. So what should clinicians tell patients about the harms and benefits of screening?
D Ilic: Clinicians should refer to the ERSPC and PLCO studies, and their various strengths and limitations. The potential benefit of screening (as reported by the ERSPC) is a reduction in prostate cancer-specific mortality (0.11 per 1,000 person-years, or one prostate cancer death averted for every 781 men invited for screening) [5]. The PLCO study (and others) would suggest no reduction in prostate cancer mortality from screening. Potential harms include overdiagnosis, implications of false-positive test results, adverse events from further testing including hematospermia, hematuria, infection, bleeding, urinary difficulties and anxiety [1, 43, 56]. Such detailed information can be difficult to convey to patients in a short period of time during a consultation – hence the need for greater use of decision aids and other patient education strategies to ensure that patients are truly informed of the benefits and risks of screening.
A Vickers and S Carlsson: You cite a review on the operating characteristics of PSA to suggest that its sensitivity is questionable. This review used the presence of prostate cancer as an endpoint. Many would argue that it is rather irrelevant whether PSA has good or bad sensitivity for prostate cancer, because most prostate cancer is indolent. The key question is whether PSA is sensitive for potentially lethal disease. What is your view on whether PSA is a good predictor for aggressive prostate cancer?
D Ilic: The PSA test is currently the frontline test that clinicians rely on with respect to diagnosing prostate cancer. However, the PSA test is problematic in distinguishing whether the cancer will become lethal in men diagnosed with localized prostate cancer. The PSA test has been demonstrated to be a poor predictor for detecting lethal prostate cancer, as evidenced by men who have been diagnosed with localized prostate cancer and managed by watchful waiting [57]. Modifying existing strategies for using the PSA test, such as interval times and thresholds for biopsy, have been proposed, although greater emphasis is currently placed on the development of new markers that could differentiate between indolent and potential lethal cancers [49].