A cohort study reporting clinical risk factors and individual risk perceptions of prostate cancer: implications for PSA testing


Correspondence: Liz Forbat, Cancer Care Research Centre, University of Stirling, Stirling FK9 4LA, UK.

e-mail: elizabeth.forbat@stir.ac.uk


What's known on the subject? and What does the study add?

  • Prostate cancer has three known clinical risk factors: age, ethnicity and family history. Men's knowledge of prostate cancer is low.
  • This study demonstrates that men rely on family and friends to learn about prostate cancer and help them interpret their risk. The findings suggest the need for tailored prostate cancer education, through social networks, to encourage risk-stratified PSA testing, which will lead to earlier diagnosis for those most at risk.


  • To determine men's perceptions of their risk of developing prostate cancer.
  • To consider the implications for PSA testing based on individual risk perceptions.

Patients and Methods

  • The research adopted an embedded mixed-method design, using clinical records and a retrospective postal survey.
  • Patients (N = 474) diagnosed with prostate cancer in a two-year period (2008–2009) in Greater Glasgow were identified from pathology records.
  • In all, 458 men received a postal survey (16 deceased patients were excluded); 320 men responded (70%).


  • Analysis indicates that there is no association between known clinical risk factors and men's perceptions of their own risk.
  • Older men did not display increased perceived risk. Men with a family history of prostate cancer (11%) had no increase in their own perception of risk.
  • PSA tests are not requested by those who are at greater risk.
  • The subsample of patients who had requested a test were no more likely to have a family history of prostate cancer. They were more likely, however, to perceive themselves to be at high risk, to have friends with prostate cancer, to be affluent and to have a low grade tumour.


  • GPs need to balance men's risk perceptions in discussions about known clinical risk factors.
  • Men's knowledge of prostate cancer stems largely from interpersonal sources (such as friends/family).
  • Social networks may consequently offer an additional opportunity to increase awareness of risk-stratified testing.


Prostate cancer has three known clinical risk factors: age [1], ethnicity [2] and heredity [3]. However, men appear to show little personal awareness of these risk factors [4] and underestimate their risk of developing prostate cancer [5]. Indeed, men's knowledge of prostate cancer remains low [5, 6].

The picture is less clear for those with a family history of prostate cancer. Some studies have found that these men are likely to overestimate their risk [7, 8], whereas others have found no difference in risk perception [9, 10]. Perceived risk has been linked to the likelihood of prompt presentation [7], and lack of awareness of risk factors has been linked to delay in presentation [11, 12].

Testing should be based on individual risk rather than across the population or on symptomatic presentation [13]. Without a national prostate cancer screening programme [14-17], it is important to determine which factors lead to men being diagnosed with prostate cancer, including subjective risk-appraisals and requests for PSA tests, to better understand where under-diagnosis or over-diagnosis is likely to occur.

The research question is what factors contribute most significantly to men's perceptions of their personal risk of prostate cancer.

Patients and Methods


An embedded mixed-methods design was implemented [18], with two stages: (i) pathology records identified clinical data (PSA, Gleason and TNM); (ii) a postal survey was distributed to all men between July and October 2010 to ascertain core details of symptom profile, triggers to action, route to diagnosis and demographic information.


Patients were identified from pathology records across Greater Glasgow. The sample included all men who received a first diagnosis of prostate cancer between 1 January 2008 and 31 December 2009. In all, 566 men met the inclusion criteria. Of these men, 92 were excluded because clinical records were unobtainable, the patient had died or GPs had no ongoing record at their practice. Date of birth, postcode, PSA and Gleason and TNM scores were collected for the remaining cohort.

A postal survey was sent to all eligible men (N = 458). Deceased patients’ data are included in the analysis of the total patient cohort. The survey was returned by 320 men (70%). Survey respondents were younger (χ2 = 6.738, df = 2, P = 0.034) and less deprived (χ2 = 15.151, df = 4, P = 0.004) than non-respondents. Respondents tended to have a lower grade cancer (χ2 = 2.589, df = 1, P = 0.108) than non-responders, although this difference was not statistically significant. Table 1 outlines the demographic and clinical profile of survey respondents.

Table 1. Demographic and clinical profile of survey respondents
Characteristic Number of menPercentage of survey respondents
Mean age of respondents68.54 years (sd = 8.053)  
Total 320100%
DeprivationCategory 1 (most deprived)7222.4%
Category 24714.6%
Category 33711.5%
Category 44313.4%
Category 5 (least deprived)12038.0%
Total 320100%
Grade of diseaseIndolent or moderately invasive21065.6%
Highly aggressive10633.1%
Total 31698.7%
Ethnic originWhite, white Scottish, white British, white other31297.5%
African, Caribbean or Black20.6%
Asian, Asian Scottish, Asian British51.6%
Total 31999.7%
Marital statusMarried/partner25479.4%
Total 31799.1%
Family historyYes3410.6%
Total 31197.2%
Total sample 320100%

Following best practice in research, the study was guided throughout by a group of men with personal experience of prostate cancer. Their roles included shaping the content and wording of the survey, designing questions and interpreting the data. The survey was developed by drawing on the extant literature on risk perception [5], delay [19] and barriers/triggers to presentation [20]. Questions focused on the time prior to diagnosis. Questions elicited nominal and ordinal categorical data.

Disease Stage Definitions

Grade of disease was determined by applying the two categories (high and low) used in the STAMPEDE trial (systemic therapy in advancing or metastatic prostate cancer: evaluation of drug efficacy) [19]. Highly aggressive disease is defined as patients at high risk of prostate-specific mortality. Highly aggressive tumours must fulfil one of the following three categories: (i) at least two out of the three criteria (a) stage T3/4N0M0 (locally advanced non-metastatic) histologically confirmed prostate adenocarcinoma, (b) PSA ≥ 40 ng/mL and (c) Gleason sum score 8–10; (ii) stage TanyN+M0 or TanyNanyM+ histologically confirmed prostate adenocarcinoma; (iii) multiple sclerotic bone metastases with PSA ≥ 100 ng/mL without histological confirmation. Age was calculated at the time of the survey being posted (July 2010).

Data Analysis

Inferential statistics (χ2 tests) were used to explore hypotheses regarding statistically significant associations between variables. The significance level was set at 0.05. When performing χ2 tests, categories were combined when necessary to ensure sufficiently large expected values.

Ethical Considerations

The study received National Health Service ethical approval (Greater Glasgow and Clyde REC3; study reference 10/SO701/35). Patients were confirmed by their primary care provider to be alive, to prevent mailing surveys to relatives of deceased patients. All patients gave informed consent prior to completing surveys and contributing to interviews.


The data indicate that men do not tend to take into account defined clinical risk factors (specifically age and heredity) when considering their own personal risk of prostate cancer. Requests by patients for PSA testing were not associated with known risk factors. Total numbers of respondents reported in the data vary in each section and table, reflecting the receipt of some incomplete questionnaires.

Age and Risk Perceptions

The mean age of the total patient population sample was 69.3 years (sd = 8.240) (slightly higher than the survey respondents indicated in Table 1) with a range from 47 to 88 years. In total, 62.7% of men (n = 291) in the total patient cohort had clinically indolent/moderately invasive disease, while 37.3% of men (n = 173) had highly aggressive disease. The data indicate that older men do tend to have aggressive disease, as indicated in Fig. 1. However, there was no significant relationship between age and perceived risk of prostate cancer (χ2 = 5.014, df = 4, P = 0.286; Fig. 2). Indeed the majority of men across all age groups considered themselves as low/average risk, with only 21 men (7%) considering themselves as high risk.

Figure 1.

Relationship between age and clinical risk.

Figure 2.

Relationship between age and perception of risk of prostate cancer.

Heredity and Risk Perceptions

In all, 11% of respondents (n = 34) reported a family history of prostate cancer (father or brother); these men were slightly more likely to perceive themselves at higher risk than men without this history, but the difference was not statistically significant (χ2 = 2.447, df = 2, P = 0.294). Most men with a family history of prostate cancer thought they were at low (n = 13, 45%) or average (n = 12, 41%) risk of the disease. Few (n = 4, 14%) reported feeling at high risk (Table 2).

Table 2. Relationship between family history of prostate cancer and (pre-diagnosis) perception of risk of prostate cancer
Family history of prostate cancer prior to diagnosis?I felt my chances of getting prostate cancer wereTotal
No141 (56%)93 (37%)17 (7%)251 (100%)
Yes13 (45%)12 (41%)4 (14%)29 (100%)
Total154 (55%)105 (37.5%)21 (7.5%)280 (100%)

Interpersonal Experience and Risk Perceptions

Men who reported having a friend with prostate cancer prior to diagnosis were significantly more likely to perceive themselves at high risk of developing the disease (χ2 = 11.922, df = 2, P = 0.003). Conversely those without a friend with prostate cancer were more likely to see themselves at low risk (61.5%). Risk perception appears to be strongly associated with interpersonal exposure to prostate cancer on a social level (Table 3).

Table 3. Relationship between having a friend with prostate cancer and perception of personal risk of the disease (percentage of men with or without a friend with prostate cancer)
Did you have a friend with prostate cancer, prior to diagnosis?I felt my chances of getting prostate cancer wereTotal
No115 (61.5%)63 (33.7%)9 (4.8%)187 (100%)
Yes39 (42%)42 (45.1%)12 (12.9%)93 (100%)
Total154 (55%)105 (37.5%)21 (7.5%)280 (100%)

PSA Requests and Risk Perceptions

Table 4 outlines the associations between respondent characteristics and PSA requests. Men who requested a PSA test were significantly more likely to have indolent/moderately invasive disease (χ2 = 5.040, df = 1, P = 0.025) and be considered affluent (χ2 = 14.792, df = 4, P = 0.005). They were more likely to perceive themselves at high risk of developing prostate cancer than those who did not request the test (χ2 = 8.642, df = 2, P = 0.013).

Table 4. Characteristics of men who requested a PSA test
Characteristic Men who requested PSA, n (%)Men who did not request PSA, n (%)
  1. Not all men answered the cross-tabulated variables, resulting in slightly different total scores across rows.
Grade of diseaseIndolent or moderately invasive32 (82.1%)175 (63.9%)
Highly aggressive7 (17.9%)99 (36.1%)
Total 39274
Scottish Index of Multiple Deprivation (SIMD) categoryCategory 1 (most deprived)2 (5%)70 (25.3%)
Category 24 (10%)42 (15.2%)
Category 35 (12.5%)31 (11.2%)
Category 44 (10%)39 (14.1%)
Category 5 (most affluent)25 (62.5%)95 (34.3%)
Total 40277
Family historyYes7 (17.9%)26 (9.6%)
No32 (82.1%)244 (90.4%)
Total 39 
Friend with prostate cancerNo14 (35.9%)198 (73.3%)
Yes25 (64.1%)72 (26.7%)
Total 39270
Knowledge of prostate cancerNo12 (30.8%)187 (68.2%)
Yes27 (69.2%)87 (31.8%)
Total 39274
Perceived riskLow13 (35.1%)141 (57.6%)
Average18 (48.6%)89 (36.3%)
High6 (16.2%)15 (6.1%)
Total 37245

In line with the findings reported above, men requesting a PSA test were more likely to have a friend with prostate cancer (χ2 = 22.175, df = 1, P < 0.001) but no more likely to have a family history of the disease (χ2 = 2.472, df = 1, P = 0.116). Consequently, men requesting a PSA test were more likely to have a friend with prostate cancer, have prior knowledge of the disease and perceive themselves to be at high risk of developing prostate cancer.

Knowledge and Risk Perceptions

In total, 40% of respondents reported having knowledge of prostate cancer prior to diagnosis (n = 114/285). Men who reported having some knowledge of prostate cancer prior to diagnosis were significantly more likely to perceive themselves at high risk of the disease (χ2 = 21.300, df = 2, P < 0.001). Indeed, of the men who perceived themselves as high risk, 76.2% stated that they had prior knowledge of the disease, compared with 28% of men in the low risk perception group (Table 5).

Table 5. Relationship between knowledge of prostate cancer and perception of risk of prostate cancer prior to diagnosis (percentage of men with and without prior knowledge of the disease)
Did you have knowledge of prostate cancer prior to diagnosis?I felt my chances of getting prostate cancer wereTotal
No113 (72.0%)60 (56.1%)5 (23.8%)178 (62.5%)
Yes44 (28.0%)47 (43.9%)16 (76.2%)107 (37.5%)
Total157 (100%)107 (100%)21 (100%)285 (100%)

Men's information about prostate cancer, prior to diagnosis, rarely came from clinical experts (who accounted for only 4.6%). Rather, men were most commonly informed by members of their social and relational networks (36.1%). That is, men gained their knowledge about prostate cancer from family and friends, not healthcare practitioners. Information-seeking was often related to symptoms.

Urinary Symptoms and Risk

Men who experienced urinary symptoms were significantly more likely to report feeling at high risk of prostate cancer prior to diagnosis compared with those who did not experience urinary symptoms (χ2 = 7.040, df = 2, P = 0.030). Of men who perceived themselves to be at high risk, 90.5% had experienced urinary symptoms. Table 6 demonstrates the relationship between urinary symptoms and risk perception of developing prostate cancer.

Table 6. Relationship between urinary symptoms and risk perception
Did you have urinary symptoms prior to diagnosis?I felt my chances of getting prostate cancer wereTotal
No60 (38.2%)34 (32.1%)2 (9.5%)96 (33.8%)
Yes97 (61.8%)72 (67.9%)19 (90.5%)188 (66.2%)
Total157 (100%)106 (100%)21 (100%)284 (100%)

Although there is a lack of evidence for the link between urinary symptoms and prostate cancer [21], urinary symptoms do feature in men's reasons for seeking medical advice. Consequently, men may postpone presentation to a GP if they are asymptomatic or if symptoms are intermittent.


The data identify that men's risk perceptions of developing prostate cancer are most strongly associated with having a friend with the disease and the presence of urinary symptoms, despite the lack of clinical evidence to support such assessments [22]. Although many men had limited knowledge of prostate cancer prior to diagnosis, those with some knowledge were more likely to consider themselves to be at higher risk. PSA tests were not requested by those who are most likely to develop the disease (older or with a family history). The uptake of PSA testing occurs more frequently in affluent men. In parallel with other studies [23], men who requested a PSA test were likely to be diagnosed with an indolent or moderately invasive tumour, indicating the potential for over-treatment in this population as well as under-treatment in less affluent groups [16]. The potential for under-diagnosis of men at risk of developing the condition has international relevance, but also local significance in what is referred to as ‘the Glasgow effect’. The Glasgow effect refers to an, as yet, unexplained phenomenon characterized by this city having poorer health outcomes than anywhere else in the UK, even when adjusting for deprivation and compared with cities with equivalent deprivation [24].

The retrospective design introduces potential recall bias. However, as a comprehensive cohort study with extremely high response rates, the data provide a unique insight into men's perceptions of risk factors compared with clinical markers of their disease. Ethnicity was not investigated as an independent variable due to the homogeneity of the sample.

A PSA threshold of above 40 ng/mL was applied in this study for patients presenting with non-metastatic disease. This is in keeping with the ongoing STAMPEDE study [19] and was selected to provide a focus on the patient group with very high risk disease [25]. In our study, this very high risk non-metastatic group is not separated from those with metastatic disease. It may be prudent in a future larger study to focus on patients with high risk non-metastatic and metastatic disease as distinct subgroups.

Perceived risk appeared to be influenced primarily by experiential factors (i.e. knowing someone with prostate cancer) rather than an objective assessment of risk. Previously, Montgomery etal. [10] found no significant relationship between men having a friend with prostate cancer and increased risk perception. This contrasting finding may reflect the smaller numbers of men in the Montgomery etal. study, as well as the differing methodology. Our study asked patients to reflect on how they perceived their risk prior to diagnosis whereas Mongomery etal.'s study in the USA asked men without a diagnosis to state how at-risk they felt. Previous studies on PSA uptake have sought to determine the efficacy of population-based testing on survival, rather than examining the characteristics of those who seek such tests [26].

Objective clinical risk factors are missing from men's risk perceptions. If men perceive their risk to be high they are more likely to report to the GP and request a PSA test. This presents an inequity, as men base their judgement to seek a PSA test on the presence of urinary symptoms, knowing someone with prostate cancer and relative affluence. When making a decision regarding the suitability of a PSA test, clinicians should discuss with patients the known risk factors (age, heredity, ethnicity) but also be cognizant of the factors that contribute to men's own risk perceptions. Men at relatively higher risk can thereby be targeted for a test, while recognizing the role that risk perceptions play in symptom interpretation and attendance at the GP.

Conclusions and Clinical Implications

Men request PSA tests on the basis of their perceptions of risk. These risk perceptions are likely to be influenced by having previous knowledge of the disease and knowing someone who has been diagnosed as well as concerns about urinary symptoms. These findings reinforce the need to target prostate cancer education in ways that match men's social experiences. Men who would be classed as ‘at risk’, due to family history for example, may not present if they are not experiencing symptoms. In a future study, it would be interesting to explore the use of a ‘risk-calculator’ tool, to examine the relationship between clinical and demographic variables (such as PSA, family history and ethnicity) and patient-reported risk appraisals.

Since the significance of survival benefit resulting from population-based screening for prostate cancer remains unclear [23, 27, 28], men most likely to develop disease should be encouraged to be aware of their level of risk. These data suggest that an important source of education exists at the social level (friends, family and colleagues); such networks could be utilized to increase risk-stratified testing. These creative approaches to health promotion are likely to be of most significance for men in diverse social contexts, including those with complex combinations of risk such as ethnicity, deprivation and access to healthcare. GPs should be aware that clinical risk variables play only a small role in the reasons why men present to their GP.


This study was funded by the Prostate Cancer Charity Scotland, who had no role in the conduct of the study. Thanks to pathology colleagues, Fiona Winslow for the clinical data and Aileen Ireland and Angela Ironside for administrative support.

Conflict of Interest

None declared.