Background

Prostate cancer incidence has increased in the last two decades, due to increasing prostate specific antigen (PSA) testing and subsequent prostate biopsy [1]. Results from two large randomised controlled trials, have not definitively answered the question regarding the benefit of PSA testing on mortality [2, 3]. Despite this, the rate of PSA testing in Ireland is high [4] and the number of PSA tests performed continues to rise [5]. The majority of PSA tests originate in general practice [4, 5]. This opportunistic testing has led to an increase in prostate cancer incidence, younger age at diagnosis and a shift towards more localized disease [5, 6]. Increased prostate cancer detection has important consequences for men’s quality-of-life [7, 8]. There is no consensus between guidelines internationally on the use of PSA testing [9, 10] and PSA testing of asymptomatic men is not a national policy in Ireland [11, 12]. It is therefore important to understand factors associated with PSA testing of men in general practice.

Psychological health negatively impacts breast, cervical and colorectal cancer screening [13, 14]. Its impact on cancer screening in men [15] and on PSA testing has received little attention, however, and results have been conflicting due to small sample sizes and different measures used [16–18]. In addition, a small number of studies have recently reported that markers of healthcare utilisation influenced whether men have PSA tests and other cancer screening [17, 19, 20].

Our objective was to investigate, at the population level, associations between PSA testing and men’s psychological and physical health and their health services utilisation.

Methods

Setting

Ireland has a mixed public-private healthcare system. Approximately one-third of the population are eligible for the state-funded General Medical Services (GMS) Scheme, as determined by means-test and age [21], which entitles them to free General Practitioner (GP) and hospital visits and prescriptions for a small fee per item. GPs are reimbursed for GMS patients by the Health Services Executive. Approximately half the population have private health insurance (PHI). However, most insurance plans do not cover GP visits, and patients pay between €50 and 60 per visit.

Study population

This study population consisted of males aged ≥50 years participating in wave 1 (2009–2011) of The Irish Longitudinal Study on Aging (TILDA) [22]. TILDA is a study of the health, lifestyle and financial situation of a population-representative sample of people aged ≥50 years involving Computer Aided Personal Interview (CAPI) in participants’ homes, a Self-Completion Questionnaires (SCQ) and comprehensive health assessment (HA) in one of two health centres. Where travel to health centres was unfeasible (~10% of participants), nurses performed the HA in participant’s homes (Additional file 1). TILDA was approved by the Faculty of Health Sciences Research Ethics Committee of Trinity College Dublin. Potential participants, who were unable to give consent due to dementia or cognitive impairment, were excluded.

Outcome variable

The main outcome variable was ever having had a PSA test. Men were included if they gave a definitive answer to the CAPI question asking had they ever had “a PSA blood test to screen for prostate cancer”. Men who responded “don’t know” or declined to answer were excluded (n = 116).

Covariates

Healthcare utilisation

The self-reported healthcare utilisation variables recorded were i) number of GP visits in the previous year; eligibility for GMS [23] (yes/no), ii) cholesterol testing (ever/never), iii) influenza vaccination (ever/never), and iv) number of regular medicines (prescription/other) including chronic cardio-preventative medication, statins, and aspirin (yes/no) classified using WHO ATC Classification.

Three scales were used to measure psychological health; depression was assessed using the Centre for Epidemiologic Studies Depression (CES-D) scale (scored as 0–7 not depressed; 8–15 sub-threshold; ≥16 case-level depression [24]); anxiety was assessed using the Hospital Anxiety and Depression Scale (HADS-A): (scored as 0–7 not anxious; 8–10 borderline; ≥11 case-level anxiety [25]); global cognitive function was assessed using the Mini Mental State Examination (MMSE; scored as 26–30 normal cognitive function; 20–25 mild cognitive impairment; <20 moderate cognitive impairment [26]). Mild and moderate cognitive impairment groups were combined because of the small number of men in the latter group. Participants for whom data was unavailable were classified as “unspecified” for each of these categories [22].

Men’s overall physical health was measured by summing the number of self-reported chronic illnesses from the following list: heart attack, heart failure, angina; stroke; diabetes; hypertension; high cholesterol; lung disease; asthma; cataracts; cancer; Parkinson's disease; peptic ulcer; arthritis; osteoporosis or hip fracture. Men taking medications in the WHO-ATC category G04C were classified as having been treated for Benign Prostatic Hypertrophy (BPH). Previous cancer diagnosis (yes/no) was identified separately. A frailty score was derived within TILDA from five measurements; self-reported weight-loss of ≥4.5 kg in the year pre-interview; weakness based on grip-strength; self-reported exhaustion; gait speed; and low physical activity. Other variables associated with frailty were: self-reported arthritis, joint replacement and osteoporosis (yes/no), hip or wrist fracture (ever/never). Subjective health status variables investigated included overall self-rated health, and self-rated emotional or mental health (excellent/very good, good, fair/poor).

Socio-demographic characteristics including age, marital status, work status, smoking status, highest educational level achieved and PHI status, were recorded at the CAPI.

Statistical analysis

Univariate analyses (chi-square test, Wilcoxson rank-sum) were used to identify associations between covariates and ever having had a PSA test. Logistic regression was used to build a multivariate model of predictors of PSA testing. Analysis was conducted in two stages. Firstly, a core model was developed from socio-demographic, healthcare utilisation and health status variables previously associated with PSA testing (age, marital status, education, employment, smoking status, number of GP visits) and covariates with a p-value <0.1 in univariate analyses. Collinearity was addressed by including one of two potentially correlated variables (e.g. number of chronic illnesses, but not number of medicines). Covariates retained in the core model were: number of chronic illnesses, influenza vaccine, prior cancer diagnosis, treated for BPH, and GMS eligibility. In stage two, psychological and physical health measures were added separately to the core model, to assess their independent association with PSA testing.

Sensitivity analysis was performed to examine effects on multivariate risk estimates of PSA testing of excluding men who had a previous prostate cancer diagnosis (n = 93). Individual comorbidities were assessed for association with PSA testing in the core model, as an additional analysis.

TILDA data V 1-7-3 and STATA V 12 were used for analyses. Significance at p < 0.05 was assumed.

Results

Study population characteristics

The median age of men was 63 years (IQR 56–71, N = 3,628). 68.2% reported ever having a PSA test (Table 1). Of these men, 84.2% returned the SCQ, and HA data was available for 72.3% (Figure 1).

Table 1 Characteristics of the study population, stratified according to whether they ever or never had a PSA test (N (%), unless otherwise stated)
Full size table
Figure 1
figure 1

Flow diagram of the study population identified from the TILDA study.

Full size image

Stage 1: Core model

The core multivariate model is presented in Table 2. Ever having a cholesterol test was the factor most strongly associated with PSA testing in univariate analysis (OR = 17.0 95%CI 12.9-22.4). Therefore, to assess other independent associations with PSA testing, this variable was removed from the model. In multivariate analyses, physical health (more chronic conditions (OR = 1.11 per unit increase in conditions, 95%CI 1.05-1.19); previous cancer diagnosis (OR = 2.74, 95%CI 1.74-4.30); BPH treatment (OR = 2.66, 95%CI 1.65-4.27)), healthcare utilisation (increased number of GP visits (OR = 1.01, 95%CI 1.01-1.05; having an influenza vaccination (OR = 1.35, 95%CI 1.13-1.60)); and socio-demographic variables (higher educational attainment and being married/cohabiting compared to other marital status) were associated with increased likelihood of having PSA tests. Men were significantly less likely to have had PSA tests if they were: current smokers (OR = 0.56, 95%CI 0.45-0.69), GMS eligible (OR = 0.63, 95%CI 0.52-0.77) or were not employed (OR = 0.67, 95%CI 0.53-0.85) (Table 2).

Table 2 Univariate (OR 95% CI) and multivariate (OR 95% CI) analysis of associations between socio-demographic characteristics, healthcare utilisation, and physical health, and ever having had a PSA test (Stage 1: Core Model)
Full size table

Stage 2: associations between psychological and physical health and PSA testing

One-fifth of these men had depression, of whom 15% and 7% had sub-threshold and case-level depression, respectively. Prevalence of borderline and case-level anxiety were 16.3% and 5.4%, respectively. Men with sub-threshold depression were significantly less likely to have had a PSA test(s) (adjusted OR = 0. 79, 95%CI 0.62-0.97). Men with case-level anxiety had reduced likelihood of PSA testing in unadjusted analyses compared to non-anxious men, but this was not significant following adjustment (OR = 0.79, 95%CI 0.57-1.09). Lower self-rated emotional or mental health was associated with reduced likelihood of PSA testing in univariate analysis, but this was no longer significant in adjusted analyses (Table 3).

Table 3 Univariate (OR 95% CI) and multivariate (OR 95% CI) analysis of associations between psychological and physical health covariates, and ever having had a PSA test (Stage 2)
Full size table

Patients with a degree of cognitive impairment were significantly less likely to have had PSA tests. Those with mild-moderate cognitive impairment were less likely to have had PSA tests, compared to those with unimpaired cognition, though non-significantly (OR = 0.79, 95%CI 0.58-1.08).

Frailty was associated with reduced likelihood of PSA testing, this was significant for men who were pre-frail (adjusted OR = 0.68, 95%CI 0.56-0.83). Individual frailty measures associated with non-testing were low grip strength (OR = 0.84, 95%CI 0.69-1.02), low gait speed (OR = 0.61, 95%CI 0.43-0.86) and low levels of physical activity (OR 0.66, 95%CI 0.50-0.87) (Additional file 2). Men who reported heart attack/heart failure/angina (OR = 0.62, 95%CI 0.47-0.80), stroke (OR = 0.55, 95%CI 0.32-0.95) and lung disease (OR = 0.64, 95%CI 0.43-0.95) were significantly less likely to have had PSA tests in adjusted analyses (Additional file 3).

Exclusion of men who had a prostate cancer diagnosis (N = 93) did not affect associations between any covariates in the core model and ever having a PSA test, except previous cancer diagnosis (Additional file 4).

Discussion

PSA testing is widespread in Ireland. However, our findings suggest that men with lower self-reported physical and psychological health, including depression, anxiety, cognitive impairment and frailty were less likely to have had a PSA test, while men with very good self-reported health were more likely to have had a PSA test, in this nationally representative sample of men aged 50 years or older, after adjusting for socio-demographic factors. Increased healthcare utilisation was also associated with increased likelihood of PSA testing, however men who were eligible for free healthcare were less likely to have been tested.

These results should be interpreted with some care, as the study was cross-sectional in design; however we considered three hypotheses of health behaviour to explain these observations [27, 28]. Firstly, there is evidence of a ‘healthy user effect’ [27] whereby men taking preventative medication e.g. statins and receiving influenza vaccinations were more likely to have had PSA tests. The healthy user effect is a multidimensional concept incorporating ‘health-seeking’ tendencies, i.e. healthier patients request or accept more screening tests and have increased adherence to medications, but it also incorporates ‘health status’ i.e. the ability of patients, physically and cognitively to attend primary care and to get prescriptions filled [27]. Multi-morbidity results in polypharmacy and increased health services utilisation [29]. We found that, despite adjusting for number of GP visits, men with more chronic illnesses were more likely to have been tested, suggesting that some PSA tests can be ascribed to the ‘surveillance hypothesis’ i.e. men with coexisting conditions have more frequent contact with the healthcare system facilitating early diagnosis [20, 28]. However, while not the central focus of this paper, we found that the association between comorbidity and PSA testing depended on the coexisting disease, which was in agreement with other studies [30]. In this cohort, likelihood of PSA testing was increased in men with angina, high cholesterol, cataracts and hypertension (‘surveillance hypothesis’), but was negatively associated with frailty, cardiac diseases and stroke, suggesting that poorer physical health may distract a GP from undertaking, or offering men a PSA test, the ‘competing demand hypothesis’ [28].

The negative associations observed, in this cohort, between poorer psychological health and likelihood of PSA testing are further evidence of the ‘competing demand hypothesis.’ The impact of poor psychological health on the likelihood of men having a PSA test has received little attention. Men with sub-threshold depression were significantly less likely to have PSA tests than men who were not depressed, suggesting that somatic symptoms associated with depression may be more pertinent during healthcare visits, or that GPs may be less likely to initiate discussions about PSA testing and prostate cancer with depressed men for fear of exacerbating their condition [31]. Our findings concur with previous work which observed lower rates of breast, cervical and colorectal cancer screening among people with depression [13, 14, 32], despite increased usage of primary care services [14]. However, we found that case-level depression was not associated with PSA testing. This may be due to various reasons; men with case-level depression may be receiving management for depression and thus may be more likely to be PSA tested, consistent with the ‘surveillance hypothesis’; the number of men with case-level depression may be too small to detect significant effects; or the effect of case-level depression on PSA testing may be no longer significant when other aspects of psychological health e.g. anxiety was included in the model. In support of the latter hypothesis, Kotwal et al. observed that men with depressive symptoms were less likely to have PSA tests however, this effect was mediated by levels of perceived stress [19] which concurs with the surveillance hypothesis. While stress was not measured in this cohort, we found that case-level anxiety was associated with reduced likelihood of PSA testing in univariate, but not adjusted analysis.

Anxiety has been shown to influence PSA testing in a number of ways. A review found variations in anxiety levels across the prostate cancer continuum from screening to beyond treatment [16]. Dale et al., observed that PSA screening was associated with increased levels of anxiety in men, highest in those with high pre-dispositional anxiety, but this mostly subsided upon receipt of a normal result [16]. Furthermore, anxiety about having prostate cancer predicted both PSA testing and avoidance of screening, the latter especially among asymptomatic men with a family history of prostate cancer. To further elucidate the effect of anxiety on PSA testing Consedine et al., in a small study of 533 American men, aged 45 to 70 years, investigated the effect of three components of anxiety i.e. dispositional anxiety, prostate cancer worry, and screening fear on frequency of PSA testing (and digital rectal exams (DRE)) [18]. They concluded that cancer worry propelled men to have PSA tests, screening fear deterred men from having DRE, but not PSA tests, and trait anxiety was associated with more frequent DREs, but not with PSA testing [18]. Furthermore, a recent study observed that the association between anxiety and PSA testing is dependent on the number of GP visits [19]; men were less likely to be PSA tested if they had higher anxiety and attended their GP once, but men with higher anxiety who attended their GP more frequently were more likely to be tested. However, it is difficult to draw comparisons between studies due to different study designs, populations examined and instruments used to measure psychological health (depression, anxiety and stress) [13–16, 18, 19].

Men with increased cognitive impairment were also significantly less likely to have had PSA tests, which again may be explained by the ‘competing demands’ hypothesis. This is the first time associations between psychological health and PSA testing has been observed in men in Ireland and our findings add to the growing body of literature on the effect of psychological health on preventative health and cancer screening.

Eligibility for free healthcare (GMS eligibility) is associated with more frequent GP visits [23], however, despite adjustment for socio-demographic, health, and healthcare factors including number of GP visits, GMS eligibility was negatively associated with PSA testing, which is consistent with income-related inequality in uptake of PSA testing observed in Ireland and elsewhere [33]. This highlights the issue that in mixed public-private systems, free healthcare services does not produce equity in uptake of primary care services, and may in part explain the higher prostate cancer incidence in higher socioeconomic groups [34].

Socio-demographic factors were strong predictors of PSA testing and our findings are broadly in agreement with others [19, 20, 33]. Married men were more likely to have PSA tests possibly because their wives engage in breast and cervical cancer screening (Drummond et al. unpublished data). Odds of PSA testing were greatly reduced in current smokers, this trend concurs with findings from previous studies [35]. Smoking-related illnesses may be prioritised by GPs, the ‘competing demands hypothesis’ and/or smokers may avoid engagement with health services because they anticipate unwanted advice to quit smoking [36].

This study has several strengths. It is a large sample, representative of the population [22], with data on a wide range of variables. Standardised measures of depression, anxiety and cognitive function were used, although stress was not measured. We acknowledge several limitations; data on PSA testing was self-reported, which is subject to recall bias [37]. However, that more than two-thirds of men reported ever having a PSA test is not surprising given the level of PSA testing in Ireland has risen since the mid 1990’s [38]; by 2004, 41.2% of men aged 50 years and older had had a PSA test [6], and the number of PSA tests continues to increase [5]. Sensitive information may have been withheld e.g. use of anti-depressants; or chronic conditions misclassified and the strength of some associations with PSA testing may have been limited due to small numbers in sub-groups. A small number of men in this study (n = 93; 2.6%) had a previous prostate cancer diagnosis, and the prevalence of psychological distress (including depression and anxiety) has been shown to be elevated in men with prostate cancer in some studies [39, 40], but not others [41]. Analyses were adjusted for previous cancer diagnosis and a sensitivity analysis was conducted excluding men with a previous prostate cancer diagnosis (Additional file 4: Table S4). Finally, there is potentially residual or unmeasured confounding in the analysis for example the influence of GPs, or ‘provider effect’, on whether men were PSA tested.

Conclusions

In conclusion, this cross-sectional study provides insight into the characteristics of men who have, and have not had PSA tests in primary care. Men in poorer psychological and physical health, smokers and those eligible for free GP services were less likely to have had PSA tests while men in good overall health and those engaging in health-seeking behaviours were more likely to have been tested. These findings might be considered by physicians and policy makers in the development of public health strategies to improve the appropriateness and equality of prostate cancer detection.

Authors’ information

EMF: Irish Cancer Society Research Scholar in the Department of Pharmacology and Therapeutics TCD (2010–2013) – focusing on Prostate Cancer Pharmacoepidemiology; FJD, PhD is a research fellow/project coordinator at the National Cancer Registry; KB, PhD is an associate professor in Pharmacoepidemiology and statistician in the Department of Pharmacology and Therapeutics TCD; ITB, PhD, is a research fellow in the Department of Pharmacology and Therapeutics TCD; LS PhD, is an epidemiologist at the National Cancer Registry and adjunct professor, Department of Epidemiology and Public Health, UCC.