Cancer screening has become an obligatory appendix to all periodic and continuity care of adults. It is ritualistic and seldom questioned. It is discrete and modular and so easily audited for ‘quality’. Its impact on actual health ranges from slightly helpful to slightly harmful, and even these conclusions are usually more than slightly speculative. While, in some cases, the existing data includes the probability of an occasional life saved, all screening programmes increase many-fold the lifetime chance of confronting a false alarm and thus becoming, for a time, a cancer patient.
The intent is deceptively simple. It is assumed that solid tumours pass through an asymptomatic stage that is screen-detectable only, followed by further growth with clinical signs and symptoms. It is further assumed that most in this early stage are curable while only some in the later stage may still be. The distinction between MOST and SOME can be re-phrased as a Window of Opportunity that justifies (perhaps even mandates) screening. The great applications of these assumptions are for breast cancer in women and prostate cancer in men, in both sexes the cancers with highest incidence and second highest mortality.
In the case of men, an astounding industry has grown from these assumptions, at least in the United States. In a saner time, the early 1980s, prostate cancer was found in 60,000 (1) men each year and about 6000 prostatectomies (2) were performed. Most of this activity derived from symptomatic cases and some from digital rectal examinations. A serum protein with specificity for the prostate gland was discovered around 1980 and within a few years was being widely deployed for cancer screening. This ‘Prostate Specific Antigen’ led to a prostate cancer incidence of 200,000 (3) by 1994 and to over 55,000 radical prostatectomies that year. Even higher rates have persisted to the present time (79,875 radical prostatectomies in 2007) (4).
Over 70% of American men have had Prostate Specific Antigen (PSA) testing (5), often on a yearly basis, as part of routine primary care. After four annual tests and a cutpoint of 4.0 ng/ml, 21% will be ‘screened-in’ and referred for blind needle biopsy (6). This entails a $4000 procedure and pathology charge. About 25–30% of these biopsies will show cancer (5). Although it is argued that clinical factors and reassuring features of some positive biopsies can still derail the rush to treatment, in practice the momentum at this point is often irresistible. If age or co-morbidities contraindicate aggressive treatment why screen at all? The classic histologic guide, the Gleason Score of tumor differentiation, can predict growth potential and malignant behaviour but, sadly, only 5% of biopsies show reassuring dormancy (Gleason 2, 3, or 4) and only 20% show clear features of high risk (Gleason 8–10) (7). Most biopsies have intermediate features of prognosis and provide scant guidance.
With mammography or hemoccult, a ‘false positive’ screening test means that biopsy rules out the presence of cancer. In the special case of prostate cancer, the entire concept of false positive becomes undefinable. With or without screening, about 3% of American men die of this disease while 70% of the prostates of very old men who die of other causes harbour ‘cancer’ (5). If you look you will find. The vast majority of these histological cancers’ are harmless unless discovered and thus the sensitivity of screening tests is more nuisance than benefit. Discovery of a 1 g cancer in the breast or colon is usually considered a good day’s work, while finding a small prostate cancer might be met with dismay or even remorse for having looked.
This conclusion is not obvious at first glance and, in fact, several outcomes seem to validate the enterprise:
- • PSA screening finds small, early-stage tumours.
- • Prostate cancer mortality is much lower among screen-detected cases than in historic controls or in concurrent clinical cases.
- • The 5-year survival for prostate cancer in 1980 was 70% and is over 95% today (8,9).
- • The vast majority of men who have undergone this screen-driven treatment are apparently free of cancer, and most with long follow up have since died of other causes.
- • America is filling up with older men (a remarkable number of them physicians) who believe they are only alive because of PSA screening and radical treatment.
What else do we need to know?
It turns out that 5-year survival is the wrong metric. It is, of course, the survival of cases diagnosed and entered into follow-up. As just noted, this denominator (‘cases’) has increased by over 200% as a dormant reservoir of millions of silent tumours is mined by PSA screening and registered for survival analysis. This stuffing of the denominator with non-clinical cases (‘Over-diagnosis’) will raise the 5-year survival even if none of these men is cured or even treated.
For exactly the same reason, the testimonials of satisfied customers are also worthless. Men with permanently dormant tumours will remain free of clinical cancer after unnecessary prostatectomy or radiation therapy. Understandably, having elected and endured this ordeal (and often permanent impotence and incontinence as well), they want to believe in the benefit. In addition, of course, the professional communities that offer these treatments (and the primary care/prevention vendors who enable the recruitment) endorse and encourage this belief.
Why do the screen-detected cancers do better than current or past clinical cases? Apart from the Over-diagnosis just described, there are three classic biases that plague most screening series. First is Volunteer Bias: people who seek out screening are health conscious and have, as a group, unusual success in avoiding or resolving illness of all types. Second is Lead-time Bias: screening finds prostate cancer 5–11 years on average before the advent of clinical symptoms (10). Even if the ultimate date of death is unaltered, survival from diagnosis is much longer. Nonetheless, most people choose interventions in hopes of dying later, not beginning to die sooner!
The third bias imputing a deceptive advantage to screen-detection is called Length Bias. Prostate cancer is famously indolent but some cases progress much faster than others. Screening is periodic, usually annual. Men with aggressive tumours pass from undetectable to metastatic before or between screening opportunities, while the more typical, indolent cases are screen-detectable for several years before spreading. Thus screening yields mostly slow tumours that are inherently curable while symptomatic cases are self-selected for lethality.
All of this essay so far has been theory and explains how screening might seem helpful without being so. These false methods of observation do not, however, prove the absence of benefit. Before we turn to the proper studies, one final dataset must be confronted: the age-adjusted death rate for prostate cancer is falling. Per 100,000 US men, this rate rose from 30 in 1980 to 38 in 1993, and has since fallen steadily to 25 in 2005 (5). The chronology is compatible with a benefit of widespread screening but at least five other contemporaneous factors could be in play:
- • Attribution Bias: The data derive from death certificates, and the cause of death in old men is not always certain. When prostate cancer has been diagnosed, it becomes a plausible option that will sometimes be chosen. This likely explains the sharp rise in the rate through the early screening years. More recently, belief in the value of radical treatments received may have diverted this choice away from prostate cancer (11).
- • Although seldom credited with cure of metastatic disease, hormone and chemo-therapy have greatly improved and patients are living longer with their incurable cancers. When old men live longer, they have opportunities to die of other causes.
- • Treatment of these other causes (especially heart disease) has improved dramatically. Even when prostate cancer remains the ultimate designated cause, the deaths are occurring later and the age-adjusted mortality drops.
- • What has declined is the total prostate cancer death rate. We have fairly good evidence that prostatectomy is beneficial for men with symptomatic, locally-advanced cancer (12). We have no such evidence for the asymptomatic tumours found by PSA screening.
- • Finally, out of the blue sky, comes the observation that the use of Statins for coronary prevention has a dramatic, and still unexplained, benefit in reducing progression of prostate cancer to advanced or metastatic disease (13). If this holds up it would explain both the time and place of the greatest drop in mortality – after 1995 and in the US.
The proper metric for evaluating a cancer screening programme is population-based, disease-specific mortality in a randomised trial of screening with unscreened controls. Such trials have been mounted over the past four decades for mammography in women and we have flawed but fair evidence of modest benefits (and very high costs). PSA screening has been widely promoted and deployed with no valid evidence of benefit and in the face of certain costs and distressing harms. Finally in March 2009, we had advanced interim reports of two long-awaited Randomized Controlled Trials (14,15) and we now have definitive evidence that the harms of PSA screening far exceed the benefits.
These randomised trials of PSA screening were launched in 1991 in both America and Europe; the cited reports (14,15) follow-up most subjects to 7 -- 9 years. The US trial (14) was smaller (38,000 screened, 38,000 controls) and plagued by massive ‘contamination’, meaning that 53% of the control group had independently received at least one PSA screen. Protocol adherence in the group assigned to screening was 82% so there was a real but modest difference. More cancers were found and treated in the screened group yet the outcome was a non-significant increase in prostate cancer mortality with a relative risk of 1.13 (95% CI 0.75–1.70) compared with controls.
The European Trial (15) was actually seven similar but not identical protocols mounted in seven countries and pooled. A total of 72,900 men were screened and 89,000 served as controls. Contamination of the European control group was not reported but is believed to be much lower than in the US (6% vs. 52%) (16). Cancer was found in 8.2% of the screened men vs. 4.8% of controls, and this time there was a significant relative reduction of prostate cancer death (RR = 0.80; CI 0.65–0.98). One life was saved at 9 years for every 1410 men screened (absolute risk reduction = 0.71/1000). Although staggering, this statistic is not far from the results of mammography or colon cancer screening. Cancer screening is profoundly inefficient and busy doctors who screen their entire practice panels according to standard guidelines will save very few lives over their careers.
There was a much more serious problem in the results of the purportedly positive European Trial. Compared with controls, 48 additional men per 1000 had to be treated for cancer as a result of PSA screening to save one life after 9 years. This number-needed-to-treat is seven times higher than with mammography for breast cancer (17). Considering the radical and morbid nature of these treatments, the advanced age of most candidates, and that this large ordeal must be incurred by 48 men electively at the outset for the prospect of saving one man 9 years later, the entire enterprise is exposed as absurd.
Debunking an established programme of cancer screening is a tough sell. Over 25,000 men in America will die of prostate cancer in 2010 and far more will do so around the world. A sceptical conclusion about PSA screening will not save these men; the problem remains and the fear abides. It therefore remains a compelling direction for future research to address this need. But for the present time and the present methods in use, it is time to call a halt. It is time to accept that the epidemic of real prostate cancer has been over-shadowed by a greater epidemic of ineffective and destructive screening. It is time to call a halt.
The first responses from the expert communities have been mixed. Prominent urologists have angrily denounced the American study and seized on the European results (exaggerating the benefit and ignoring the harms) as a validation of screening (18,19). The American Urological Association has moved into a full defensive crouch by actually endorsing an expansion of PSA screening, starting at age 40 for all men and dropping the cutpoint to 2.5 (20) [which raises the recall rate for biopsy from 21% to 37% after four annual screens (6)]. On the other hand, the American Cancer Society, which has traditionally blessed all screening tests, is finally expressing misgivings about this one (16,21). Given the recent public and political outrage that ensued when the US Preventive Services Taskforce suggested more restrictive guidelines on use of mammography (22), we must expect similar reactions as the disappointing facts about PSA screening move into the policy arena. Stay tuned…