Because of the uncertainty regarding the long-term risk of a prostate cancer that is found to be low grade on a prostate biopsy, a man with newly diagnosed prostate cancer must weigh the risks of harm from disease without treatment, against the risks of treatment side-effects. Both radiation therapy  and surgery  reduce prostate cancer mortality by ≈50% a decade after diagnosis for men with non-screen-detected prostate cancers that are localized to locally advanced. Accounting for the lead time with PSA testing, results from the SPCG-4  would suggest that there is minimum risk of death from prostate cancer in 10–15 years without treatment for a man with low-risk disease. However, given the absence of longer term data from randomized trials to show the safety of a surveillance approach, men with more than an estimated 10–15 year life expectancy and low-risk disease should be told that active surveillance is an experimental approach that may result in a lost opportunity for disease control. For those men who are older with associated disease comorbidity that limits the remaining years of life, active surveillance could be considered the best management .
A lost opportunity for effective disease control in men with low-risk disease on surveillance may occur due to the under estimation of cancer grade on prostate biopsies and/or the progression of a low-grade cancer to one of higher grade. It has been estimated that ≈25% of men classified at diagnosis with low-risk prostate cancer will be found to have higher grade disease at surgery , and that ‘true’ disease progression from low to high grade is unusual, at least in the short term . An unanswered question is the extent, if any, to which a man on surveillance who undergoes delayed intervention risks losing the opportunity for control of disease. Non-randomized data comparing men with low-risk disease who have undergone immediate vs deferred treatment, and decision analyses, have been used to inform management in the absence of randomized trials.
Stattin et al.  evaluated the outcomes of men at an average age of 61–65 years with low-risk prostate cancers who underwent surveillance and curative intervention in a large national cancer registry. The overall 10-year prostate cancer-specific mortality was <3% for men with low-risk disease managed initially with surveillance. The absolute difference in prostate cancer-specific mortality at 10 years for those managed with surveillance compared with surgery was 2%. This could be compared to a 5% absolute difference in the SPCG-4 that compared no treatment with surgery . The authors concluded that surveillance may be a suitable option for many men with low-risk prostate cancer, a stance that is consistent with recent guidelines .
An evaluation of >3000 men with a mean age of 68 years in the Health Professionals Follow-up Study, showed that 10% of men overall deferred treatment, and that at 8 years 51% remained untreated . Of those treated, on average 4 years elapsed before treatment. At a median follow-up of 8–9 years, there were no differences in the rates of metastatic disease or prostate cancer deaths when comparing those who deferred treatment with those who were treated initially. The prognostic risk category (low, intermediate, high) was strongly predictive of the lethal phenotype. When compared with those with low-risk disease, those with intermediate- and high-risk disease were three- and six-fold more likely to die of prostate cancer, respectively. When comparing deferred and immediate treatment, there were no differences in prostate-specific outcomes for men with low- or intermediate-risk disease. Among men with low-risk prostate cancer, prostate cancer metastases occurred in seven of 139 (5%) and 33 of 1252 (3%) of the men in the deferred and treatment groups, respectively; death from prostate cancer occurred in three of 139 (2%) and nine of 1252 (1%) of the men in the deferred and treatment groups, respectively.
In the longest prospective contemporary study of active surveillance for low- to intermediate-risk disease, Klotz et al.  followed 450 men (median age 70 years) with low- and intermediate-risk prostate cancer for a median of 7 years. In all, 17% of the subjects had a Gleason score of 3+4, 83% were Gleason ≤6, and 71% were considered low risk by D'Amico criteria (Table 1) [10,13–15]. Thus, one in three men had intermediate-risk disease. The 10-year actuarial prostate cancer-specific survival was 97%. Five deaths have occurred in this cohort between 4 and 10 years after diagnosis, and all deaths occurred in men who were re-classified as higher risk disease during follow-up. By contrast, in the SPGS-4 , untreated men aged >65 years had a 12-year prostate cancer-specific survival of ≈88% that did not differ from those that underwent surgical treatment. It could be argued that deaths in the Klotz et al. study  occurred in men who had advanced disease to begin with, and that surveillance did not compromise length of life . Alternatively, it could be argued that deaths occurred in men who were not good candidates for surveillance, and that earlier intervention would have prolonged life. Regardless, there is a definite, but as yet unquantifiable, long-term risk to men who choose surveillance as a management option for low-risk disease, regardless of how indolent the disease appears on a prostate biopsy.
Many centres are gaining experience with active surveillance for low-risk prostate cancers, although enrolment criteria and triggers for intervention differ. In a multi-institutional study  that included the University of Miami, University of British Columbia, Memorial Sloan Kettering Cancer Center, and the Cleveland Clinic, 262 patients with low-risk disease were on surveillance for a median of 29 months. The 5-year probability of remaining on surveillance was 75%, a figure that is very consistent across many studies [51,52]. Comparisons of pathological outcomes among men who undergo immediate vs delayed intervention suggest that in the short term, the window of opportunity for cure is not lost [22,53].
Computer simulations have been used to evaluate long-term outcomes in the absence of randomized trials. Hayes et al.  conducted a decision analysis to evaluate the quality-adjusted life expectancy (QALE) of men aged 65 years managed with surveillance and curative intervention strategies (brachytherapy, surgery, external beam radiotherapy). Active surveillance was associated with the longest QALE and surgery the shortest, but the results were highly dependent on a man's preferences with respect to living with cancer and having it treated. The authors concluded that for a man aged 65 years with low-risk prostate cancer, active surveillance is a reasonable initial approach.
Liu et al. , recently performed a decision analysis that evaluated the outcomes of men with low-risk prostate cancer managed with surveillance (the least common approach) and surgery (the most common approach), including the ratio of side-effects (cost) per additional year of life gained (incremental cost-effectiveness ratio). A man aged 67 years in average health with low-risk prostate cancer, would experience ≈ 10 years of side-effects for each additional year of life gained, under the most optimistic assumptions regarding post-surgical erectile dysfunction and incontinence. This ‘cost’ in terms of side-effects per year of life gained, reflects the low probability of adding years to life with surgery for most men aged >65 years with low-risk cancer. Management options with lower side-effect profiles (e.g. focal therapy) might be associated with a lower cost.
In summary, the choice of curative intervention or active surveillance for men with low-risk prostate cancer should primarily depend on age, health status, and patient preferences with respect to living with cancer, and the possible side-effects of curative management. It is possible to explore patient preferences for living with cancer and side-effects of treatment as part of shared decision making. However, the limitations of correctly classifying a patient's long-term risk of disease progression and predicting life expectancy, together with the absence of randomized trials showing the safety of active surveillance in the long term, suggest that surveillance should be considered investigational for men who have a life expectancy that extends beyond 10–15 years.
The category of very-low-risk prostate cancer is now recognised as defining a risk group that differs from the low-risk category in terms of pathological outcomes at surgery, progression risk without treatment, and recommendations for management [15,56]. The criteria for defining very-low-risk disease were first described by Epstein et al. , among men with stage T1c (impalpable) prostate cancer (Table 1). The criteria include PSA density, which has been confirmed to be a predictor of future re-classification to a higher risk category, including high-grade disease [51,57,58]. These criteria have been used by some investigators to define a subset of men for whom surveillance may be safest.
When compared with men with low-risk disease, those with very-low-risk disease are less likely to have adverse features at the time of RP, and less likely to experience biochemical recurrence after treatment [56,59]. For example, when comparing the outcomes of 769 men in the Johns Hopkins Active Surveillance Program, who did and did not meet the criteria for very-low-risk disease, those with very-low-risk disease were significantly less likely to be re-classified to a higher risk category and to undergo treatment . In addition, the PSA-free outcomes of patients enrolled in the Johns Hopkins Active Surveillance Program who underwent delayed intervention, 80% of whom had very-low-risk disease at diagnosis , appear more favourable when compared with biochemical-free outcomes of men meeting low-risk criteria who undergo delayed or immediate intervention [22,60], although other uncontrolled factors could explain these intermediate results.
In the Toronto experience with surveillance , 71% of 450 men were considered low risk by D'Amico criteria, and follow-up was available for 110 men that underwent curative intervention. The 5-year biochemical progression-free survival was 62% for those undergoing RP and 43% for those undergoing radiation. By comparison, among 192 men on surveillance who underwent delayed treatment at a median of 2 years after diagnosis in the Johns Hopkins experience with very-low-risk disease, the 5-year biochemical progression-free survival was 96% for those undergoing surgery and 75% for those undergoing radiation . These data suggest that carefully selected men with very-low-risk disease will have favourable outcomes when deferred treatment is required after initial management with surveillance.
Tosoian et al. , recently reported the outcomes of 769 men in the Johns Hopkins Active Surveillance Program. The actuarial rate of biopsy re-classification to a Gleason score >6 on annual surveillance biopsies at 10 years was 30%. In this programme, if a man had biopsy re-classification to a high-grade cancer (Gleason score >6) and underwent surgery, the proportion with a Gleason score of 3+4 was 32% and Gleason score 4+3 was 44% on final surgical pathology. The 15-year cancer-specific mortality after surgery for men with Gleason score 3+4 and 4+3 was 4.2% to 6.5%, and 6.6% to 11%, respectively  in a multi-institutional study. Thus, a man with a very-low-risk prostate cancer entering an active surveillance programme who chose surgery if biopsy re-classification occurred, would have over a 10-year period, a 10% risk (30% × 32%) of having a Gleason score 3+4 on surgical pathology and a 15-year risk of a prostate cancer death of <1% (10% × 4.2–6.5%) postoperatively; and over 10 years a 13% risk (44% × 30%) of a Gleason score 4+3 on surgical pathology, and a 15-year risk postoperatively of a prostate cancer death of ≈1% (13% × 6.6–11%). By comparison, the 15-year risk of death from another cause for a man aged 65 years would be ≈40% . This estimate, based on biopsy re-classification triggering treatment in an active surveillance programme, and contemporary surgical outcomes by cancer grade, is consistent with modelled data from Parker et al. . For most men aged≥65 years with very-low-risk disease, curative intervention would appear to be unnecessary.
Although long-term data are not available to compare the outcomes of men on surveillance with very-low-risk and low-risk prostate cancer, the NCCN  has recommended that men with very-low-risk disease and an estimated life expectancy of <20 years, be managed with surveillance rather than immediate curative intervention.