Communicating prostate cancer risk: what should we be telling our patients?

While we wait for definitive evidence of the effectiveness of screening and treatment strategies for prostate cancer, men are asking for guidance on whether they should be screened for this cancer. Personal vulnerability and perceived seriousness of the disease influence this decision,1 and population-based risks are often quoted in decision-support material for this purpose.2 We suggest that the oft-used statement “men are more likely to die with prostate cancer than from prostate cancer” may be incorrect for men diagnosed in their 50s and 60s, and that doctors need to individualise the risk of prostate cancer for their patients.

We use Australian data and published reports to show that the use of routine population risk statistics can be misleading, and suggest methods to make this risk information more patient specific. We do not address risks associated with treatment choices, nor the best methods of presenting risks.3,4

There is considerable debate about whether screening for prostate cancer is beneficial and whether opportunistic screening should be offered.5,6 The current clinical advice in Australia is not to recommend population-based screening of asymptomatic men for prostate cancer.7-9 However, most guidelines advocate that men should be able to access screening so long as they are fully informed of the potential risks, benefits and uncertainties of investigations and treatment. A man’s perception of his personal risk and vulnerability to prostate cancer has been shown to influence his decision about whether to be tested.1 In addition, because of short- and long-term implications for quality of life, men need to be supported to apply their own personal values or preferences in relation to those risks when they decide about testing (see Box 1).

Taking standard population risk estimates (Box 2) at face value, it seems reasonable to suggest to patients that being diagnosed with prostate cancer is very rare among men younger than 50 and that mortality is very rare among men younger than 60 years. Men are much more likely to die from heart disease, and to a lesser extent lung cancer, than from prostate cancer. However, there are a number of reasons why routinely available population risk information may give an incomplete perspective.

The relative population risk of conditions varies depending on the age range examined. When presenting risk estimates, relevant age ranges need to be used. For example, a 50-year old man presented with a risk table for men aged 0–75 would note that that colorectal cancer mortality was twice as likely as death due to prostate cancer. However, for a man aged 50 who is interested in assessing his future risk of dying from prostate cancer, the risk for men younger than 50 is not relevant. Including it could result in a spuriously low estimate of population risk for prostate cancer relative to other conditions. Similarly, a man currently aged 50 has at least 30 years of additional life expectancy.15 Truncating risk estimates at 75 years would also underestimate his true risk of premature mortality.

Apart from age, other factors increase prostate cancer incidence risk. Having a first-degree relative (ie, brother or father) with prostate cancer diagnosed at an early age (under 60 years) at least doubles a man’s risk of being diagnosed with prostate cancer.16 The risk increase is highest in relatives of men diagnosed before age 60, and decreases with increasing age of the affected relative.16 These estimates are primarily based on invasive prostate cancers,16 and so cannot be explained by increased likelihood of relatives having prostate-specific antigen (PSA) tests. Although there is a risk increase for African Americans,17 there is no evidence of a similar increase in risk for Indigenous Australian men.18

We also know that some men are more likely to get certain conditions because they are exposed to known risk factors for that disease. For example, latest estimates suggest that a current smoker is 6 to 15 times more likely to develop lung cancer than a non-smoker.19 Combining this with at least a twofold increase in prostate cancer diagnosis risk for a man with a family history of prostate cancer,16 we could reasonably suggest that a non-smoking man with a family history of prostate cancer would be at a greater risk of dying from prostate cancer than of lung cancer. This is a much different scenario to that implied earlier (Box 2).

There are also unknown factors. The wide international variability in incidence rates of prostate cancer,17 and increased rates among second-generation Japanese migrants to the United States compared with Japanese men living in Japan20 suggest there may also be environmental or dietary exposures that influence prostate cancer risk. Unfortunately, it is currently unclear what specific factors are involved, and the level to which they affect an individual’s risk.21

Population mortality risk estimates do not distinguish between men who already have a diagnosis of prostate cancer and those who do not. The risk of dying from prostate cancer depends on the stage at which the cancer was diagnosed,22 and the Gleason score.23 Unfortunately, these prognostic indicators are not routinely collected in Australia and it is not clear how international data can be generalised to Australia. Combining all prostate cancers (regardless of stage and Gleason score), 5-year relative survival from prostate cancer has been shown to reduce slightly as age of diagnosis increases.24 However, 5-year survival data may not provide a complete picture of risk, as it has been suggested that initially indolent cancers can progress into metastatic disease in the long term.25

The implications of a diagnosis of prostate cancer among men in their 50s need to include the greater average life expectancy of men in their 50s compared with men in their 70s. One way of doing this is to consider the expected proportions of men still alive at 80 years of age (which is close to the current life expectancy of men at birth15). For Queensland men diagnosed in their 50s, the projected 30-year cause-specific survival was 0.399. The 20-year cause-specific survival for men diagnosed in their 60s was 0.497, and the 10-year cause-specific survival for men diagnosed in their 70s was 0.618. Therefore, about three-fifths (60%) of men diagnosed at 50 could be expected to die a premature death (ie, before reaching 80 years of age) from prostate cancer. Corresponding percentages for premature death for men diagnosed at 60 and 70 years are 50% and 38%.

Based on Queensland data (Queensland Health, unpublished data, 2005), for each 100 men diagnosed in their 50s, about 30 had died from their prostate cancer within 10 years, and an additional eight died from other causes within the same period (Box 3). The corresponding values for each 100 men diagnosed in their 70s were 40 and 29. Therefore, even though the numbers dying from prostate cancer within 10 years were similar, men diagnosed at a later age were nearly four times more likely to have died from other causes within 10 years.

Combined, these figures highlight the observation that for a 50-year-old man, a diagnosis of prostate cancer is more likely to result in premature death from the disease than for men diagnosed in their 70s. This is due both to the longer average life expectancy of younger men and to the higher impact of other causes of death as men get older. These results suggest that the often-used statement “men are more likely to die with prostate cancer than from prostate cancer” is misleading, particularly for men diagnosed in their 50s and 60s.

Should the relatively increased risk of dying prematurely after a prostate cancer diagnosis for men in their 50s be communicated to men who are making this health decision? Views here will differ depending on the life experience and background of the individual. Some may take the view that until we have definitive evidence that PSA testing and active treatment has an effect on patient outcomes, knowledge about age-specific mortality is unimportant. However, most studies indicate that personal risk from, or vulnerability to, disease is perceived as an important factor by men making a screening decision.1,26,27 For men who are actually diagnosed with localised prostate cancer, the desire for cure and fear of cancer spread strongly influences subsequent treatment choices, and most men with localised disease choose active treatment,28 despite the lack of definitive proof that treatment extends life. Men who choose early detection are asking for the opportunity to have that choice. Access to the best possible information on personal vulnerability to disease is consistent with medicolegal requirements for providing patient information.29 We suggest this includes population-based estimates of risk of premature death given his age at diagnosis, and other relevant risk factors.

Population incidence and premature-mortality risk estimates relate to diagnosed prostate cancer only. Based on autopsy results, significant proportions of men may have undiagnosed histological prostate cancer, even though it may not have caused symptoms or death.30 It is likely that, with the advent of PSA testing, some of these latent prostate cancers are now being diagnosed. It is likely that current calculations overestimate the real incidence risk and underestimate the real premature mortality risk by including these PSA-detected latent prostate cancers. As we cannot yet distinguish between those patients who have latent cancers and those men whose prostate cancer has potential to cause mortality,31 we need to accept this limitation for current estimates.

The only study design that can definitively measure whether prostate cancer screening reduces mortality on a population basis, and at what cost in terms of financial costs and quality of life, is a randomised controlled trial. These trials are currently under way in Europe and the United States,32,33 but the first results may not be available for several years. However, the absence of certainty does not necessarily imply the absence of an effect. Screening may ultimately be shown to prevent men dying from prostate cancer, and, paradoxically, while prostate cancer is less common in younger men, it is these men who may benefit most.

It is also clear that, even if evidence of effectiveness is available, men will still face a dilemma in trading off the possible immediate quality-of-life costs of treatment if prostate cancer is detected against an uncertain longer term survival gain. Thus, a process of understanding and evaluating personal risk will be an essential component of informed choice for prostate cancer testing for some time to come.

2 Description of analytical strategy

Mortality data for Australian men were obtained from the Australian Bureau of Statistics. Data on the number of new prostate cancers diagnosed among Australian men were obtained from the Australian Institute of Health and Welfare for the period 1997 to 2001 (unpublished cancer incidence data, February 2005).11

Age-specific risks of being diagnosed with, or dying from, prostate cancer were calculated using cumulative rates (risk per 1000 men). These risks represent the risk of a man at the start of the age group moving through to the end of that age group, experiencing the current age-specific rates.

To maintain comparability between incidence and mortality, risks of dying from prostate cancer were also calculated using 1997 to 2001 data (unpublished cancer mortality data, February 2005).12 However, as Australian mortality data were available to 2002,12 mortality risk differentials across a range of conditions were calculated using 1999–2002 mortality data.

The ICD-10 codes for the main causes of death considered in this analysis were heart disease (I05–I09, I11, I13, I21–I26), lung cancer (C34), colorectal cancer (C18–C20), prostate cancer (C61), stroke (I60–I69, G45, G46), chronic obstructive pulmonary disease (I27, J67, J40–J44) and accidents (V01–X59).

Additional Queensland data looking at cause-specific survival were obtained for prostate cancers diagnosed between 1982 and 2000, with follow-up to 2002 (Queensland Health, unpublished data, 2005). Because of a perceived difficulty in establishing the cause of death among the oldest of the elderly,13 we limited this analysis to men aged 50–79 years. The 10-year cause-specific survival estimates were calculated using PROC LIFETEST in SAS.14 Extrapolations of the cause-specific survival curves (to a maximum of 30 years after diagnosis) were made using exponential regression models.

Population risk estimates

Current population risk estimates of Australian men being diagnosed with prostate cancer or dying from prostate cancer are shown below. For all ages, the risk of dying from prostate cancer is less than the risk of being diagnosed with prostate cancer, while both the diagnosis risk and mortality risk increase with age.

   Population risk of death for 50-year-old Australian men to age 79 by leading causes of death (1999–2002

Population risks of prostate cancer diagnosis and death, Australia 1997–2001 (per 1000 men)

Age group (years)	Risk of diagnosis	Risk of diagnosis
< 40	< 0.1	< 0.01
40–49	1	< 0.1
50–59	12	1
60–69	45	5
70–79	79	21
80–89	105	61
0–79	137	27
50–79	136	27