To assess the rate of prostate-specific antigen (PSA) testing for prostate cancer in general practice in asymptomatic and symptomatic patients.
To assess the rate of prostate-specific antigen (PSA) testing for prostate cancer in general practice in asymptomatic and symptomatic patients.
The cross-sectional study took place in England and Wales, was population-based and covered 469 159 men aged 45–84 years. Pathology data on PSA tests requested between 19 November 1999 and 31 May 2002 by general practitioners (GPs) were provided by 28 pathology laboratories. The practices recorded reasons for the tests between 1 December 2001 and 31 May 2002. In all, 391 practices in which all GP partners participated were included in the analyses.
The overall annual rate of testing in men with no previous diagnosis of prostate cancer was estimated to be 6%, of which the annual rates of asymptomatic, symptomatic and re-testing were 2.0%, 2.8% and 1.2%, respectively, after adjusting for missing values. The rate decreased with increasing social deprivation, and with increasing proportions of black and Asian populations. The overall rate of PSA testing increased significantly from 1999 to 2002.
If the recommendations of the National Health Service Prostate Cancer Risk Management Programme were applied, 14% of asymptomatic tests and 23% of symptomatic tests would have led to a referral. As the rate of PSA testing is increasing and there are uncertainties about the benefit of screening, the workload and costs in general practice and hospitals should be monitored.
National Health Service
Prostate Cancer Risk Management Programme
British United Provident Association
National External Quality Assessment Service.
Despite the lack of evidence of the effectiveness of screening in reducing mortality from prostate cancer , testing for PSA has been increasing in the UK [2,3] and will have contributed to the rising incidence of prostate cancer. The contribution of screening to this trend is unknown. Previous studies in the UK have reported on the proportion of tests for screening submitted to laboratories, but the data were not population-based [4,5]. Future demand for screening, and the effect that this will have on the workload in the National Health Service (NHS), is uncertain. Concerns about the amount of screening by PSA in England and Wales led to the development of the NHS Prostate Cancer Risk Management Programme (PCRMP, http://www.cancerscreening.nhs.uk), which provides information to help men make an informed choice about whether or not to proceed with a PSA test. This marked a change in NHS policy to allow asymptomatic testing and it included guidance to GPs on age-related cut off levels for PSA testing in men being screened, i.e. ≥ 3 ng/mL in those aged 50–59 years, ≥ 4 ng/mL in those aged 60–69 and >5 ng/mL in those aged ≥ 70.
The present study was undertaken to investigate the extent of PSA testing in general practice in England and Wales, and was conducted with the help of NHS biochemistry and pathology laboratories, and the British United Provident Association (BUPA), which provides private medical insurance. Of particular interest is the amount of testing for screening purposes both in general practice and in private healthcare. However, as there are no population-based data on private testing in the UK, it was not possible to study rates of private testing or to study regional variation.
To select a random sample of eligible pathology laboratories within each region, a questionnaire was sent to 210 laboratories in England and Wales registered for measuring PSA with the UK National External Quality Assessment Service (NEQAS) . From the laboratory databases, general practices were identified which regularly submitted pathology requests (not just for PSA) and were likely to use the laboratories for all their requests. A short questionnaire was sent to the principal partner of each practice about use of the laboratory, and signed consent was obtained from partners who agreed to participate. No financial incentive was provided because the PSA tests were taken only as part of routine care. The aim was to recruit sufficient practices to provide a population of 90 000 men aged ≥ 40 years per region, to compare a potential difference of 1.5% vs 1.85% in the rate of screening after 6 months.
Data were collected prospectively and retrospectively. The GPs were asked to record data on the reason for each PSA test taken between 1 December 2001 and 31 May 2002 using a standardized ‘yellow sticker’ on their routine pathology forms, the options being: asymptomatic patient/screening request; symptomatic patient with no previous diagnosis of prostate cancer or BPH; patient re-tested as diagnosis uncertain; patient being treated for prostate cancer.
The request forms were processed by the laboratories, and anonymized data downloaded to the research unit. Missing data on dates of birth and reason for the test were requested from GPs. For the retrospective period 19 November 1999 to 18 November 2001 there was no complete or standardized information on the reason for each test routinely recorded by the laboratories.
The male population in each practice in 5-year age groups was estimated by applying the proportions for the relevant region for 2001, obtained from the Office for National Statistics, to the total male population registered with each practice, obtained from the Royal College of General Practitioners Census 2000 (Department of Health).
The National Database for Primary Care Groups and Trusts (http://www.primary-care-db.org.uk/) provides the Townsend social deprivation score  for each primary care group in England, as well as the proportions of different ethnic groups at ward level using data from the 2001 Census. Each English general practice in the study was linked to a primary-care group and a ward using postcodes. Black Caribbean, Black African and other black groups were combined for the main analyses. Asian populations comprised Indian, Pakistani, Bangladeshi, Chinese and other Asian populations. There were few practices with a large percentage of Irish so this ethnic group was not considered separately in the analyses.
There are no population-based data available on private testing. Data were obtained from BUPA on PSA tests recorded at 37 laboratories in England and Wales, so that the characteristics of PSA testing could be compared between private and NHS sources. Analyses were restricted to 7389 records associated with a health screening check; 64% of a total of 11 611 records. The computerized data included age/date of birth, date of PSA test, test result and laboratory.
The rate of PSA testing was studied by reason for the test, age, region, ethnicity and deprivation. The likelihood ratio test was used to test for the significance of variables in the tables. The main analyses were restricted to men aged 45–84 years, as most tests were in this age group, and to practices where all partners participated, as the population figures were linked to practices and not individual partners, and the proportions of populations covered in partially participating practices could not be estimated accurately. Poisson regression was used to study the relation between the rate of PSA testing, age, region, ethnicity (square root of the percentages of black populations and Asians), quintiles of the deprivation score and 6-month periods. In the prospective period the rate of testing was analysed separately for asymptomatic and symptomatic testing. The study had full ethical approval (South Thames MREC 01/1/25).
Replies were received from 127 of 210 laboratories; of these 127, 26 were refusers and 49 were ineligible (45 lacking data, two only doing private testing, one taking part in the ProtecT study which involves PSA testing, and one refusing to accept PSA tests for screening). From the remaining 52, 28 laboratories stratified by region were selected at random. Pathology requests (not just for PSA) were regularly submitted to these laboratories by 1397 general practices. At least one partner per practice agreed to take part in 443 of the practices (32% of 1397 sent invitations, although not all of these would have been eligible if they did not use the laboratory for all their PSA requests. Thus the response rate is likely to be higher than 32%). The response varied by region from 12% in London to 52% in the South-west. Of these, there were 391 (88%) practices where all partners took part. Of all 1706 partners at the 443 participating practices, 1573 (92%) contributed data.
For PSA tests between 1 December 2001 and 31 May 2002, 16 620 records (90%) of a total of 18 545 had complete data on age, sex, date of PSA test, practice code and PSA value. Of these, 14 684 (88%) had a code for the reason for the test. After a follow-up, similar levels of completeness were found in the retrospective period, excluding the reason for the test.
The practices in this study under-represented single-handed practices and represented a more affluent population than practices in the whole of England (Table 1). The data on ethnicity suggest that the distributions in the participating practices were not very different from those for England.
|Characteristic||Participating practices||England and Wales|
|Proportion of practices with||21 (81/391)||30 (2847/9538)*|
|one partner, % (n/N)|
|Distribution of practices by:|
|(values at points in the distribution)|
|percentage of black populations†:|
|percentage of Asian populations†:|
The rates of asymptomatic and symptomatic testing, re-testing and testing to monitor prostate cancer are shown in (Table 2), giving annual rates of testing in these groups of 1.6%, 2.2%, 1.0% and 0.8%, respectively. The annual rate of testing with ‘reason not known’ was 1.4%.
|Age, years||Reason for PSA test, % (no. of tests)||Male population|
|45–54||0.4 (660)||0.4 (748)||0.1 (116)||0.01 (25)||0.3 (439)||1.2 (1988)||169 515|
|55–69||1.0 (1989)||1.3 (2580)||0.5 (940)||0.3 (520)||0.8 (1541)||3.9 (7570)||192 115|
|70–74||1.1 (522)||1.8 (851)||1.1 (542)||0.8 (394)||1.2 (572)||6.0 (2881)||47 878|
|75–84||1.0 (586)||2.0 (1200)||1.2 (701)||1.4 (863)||1.4 (831)||7.0 (4181)||59 651|
|Total||0.8 (3757)||1.1 (5379)||0.5 (2299)||0.4 (1802)||0.7 (3383)||3.5 (16620)||469 159|
Regression analyses showed that the overall rate of testing increased significantly with age, decreased with increasing deprivation (both P < 0.001), proportions of black populations (P < 0.05) and proportions of Asian populations (P < 0.001), and varied significantly among regions (P < 0.001). The rate of asymptomatic testing showed similar associations with these variables, but showed less increase with age.
The rate of testing by age and PSA level was used to investigate the likely rate of referral of asymptomatic men using the cut off levels recommended in the NHS PCRMP (Table 3). Referrals would, in theory, have been made in 13.9% of all asymptomatic tests and 22.6% of symptomatic tests, i.e. 6-monthly rates of 0.1% and 0.3% of man aged 50–84 years, respectively.
|Age, years||PSA level, ng/mL, % (no. of tests)||Total|
|50–59||0.6 (976)||0.03 (41)||0.02 (25)||0.03 (46)||0.7 (1088)|
|60–69||0.9 (1051)||0.09 (111)||0.06 (68)||0.11 (132)||1.2 (1362)|
|70–84||0.7 (723)||0.12 (124)||0.07 (79)||0.17 (182)||1.0 (1108)|
|Total||0.7 (2750)||0.1 (276)||0.04 (172)||0.1 (360)||0.9 (3558)|
|50–59||0.6 (1057)||0.04 (73)||0.03 (52)||0.07 (108)||0.8 (1290)|
|60–69||1.1 (1250)||0.13 (149)||0.09 (112)||0.24 (281)||1.5 (1792)|
|70–84||1.1 (1144)||0.19 (201)||0.16 (172)||0.50 (534)||1.9 (2051)|
|Total||0.9 (3451)||0.1 (423)||0.1 (336)||0.2 (923)||1.3 (5133)|
The distributions of age and PSA level were compared by source (NHS and BUPA) for those tests reported to be in asymptomatic men (NHS) and in men screened (BUPA). In the BUPA records, 85% of all tests were in 45–64-year-olds and 15% in 65–84-year-olds. In contrast, in the NHS records the proportions were similar in both age groups, at 52% and 48%, respectively. The proportion of tests with a total PSA of <2 ng/mL was higher in men screened by BUPA (78%) than in asymptomatic men tested within the NHS (66%). This difference was more marked in men aged 45–64 years than in men aged 65–84 years.
Figure 1 shows the rate of PSA testing by age and 6-month period from 19 November 1999. The rate increased with age and time (both P < 0.001). The overall 6-monthly rates of PSA testing in the participating practices increased from 19 November 1999 to 18 November 2001, from 2.7% to 3.6% of men aged 45–84 years. There was little change between the 6-month period immediately before the prospective period and the prospective period itself.
By applying the percentage distribution of ‘reason for test’ to the ‘unknown’ group, and adding these numbers to the groups with ‘reason known’, the revised annual estimates of rates of asymptomatic and symptomatic testing, re-testing and prostate cancer monitoring are 2.0%, 2.8%, 1.2% and 1.0%, respectively. (Alternatively, if it is assumed that all tests with reason unknown belong either to the asymptomatic or to the symptomatic group, the rates become 3.0%, 2.2%, 1.0% and 0.8%, respectively, or 1.6%, 3.8%, 1.0% and 0.8%, respectively). A further adjustment to the distribution of social deprivation in England leads to a rate of asymptomatic testing of 1.6% (the method of adjustment is described in Appendix 2).
The overall annual rate of testing in men aged 45–84 years with no previous diagnosis of prostate cancer was estimated to be 6.0%, of which the annual rates of asymptomatic testing, symptomatic testing and re-testing were 2.0%, 2.8% and 1.2%, respectively, after adjusting for missing values. Slightly lower rates were found after adjusting to the social deprivation distribution of England (the study practices representing a higher proportion of affluent practices, which have more testing). In men with no previous diagnosis of prostate cancer the overall annual rate of testing was 5.4%, and that for asymptomatic testing 1.6% after adjusting for missing values and the social deprivation distribution. The additional contribution from private testing is unknown but the data from BUPA showed that it would have been particularly marked in men aged <65 years. There is no doubt from the time trends in this study, and compared with previous UK studies [2,3], that the rate of PSA testing is increasing in the UK. However, there are limitations with the present data which need to be considered.
Differences among regions were not reported because of variation in the response rate of practices and laboratories by region. Overall, the age distribution of men registered at the participating practices was almost identical to the distribution of all practices in the GP census, but in terms of number of partners per practice this study had relatively fewer single-handed practices than the proportion reported nationally. The true response rate among eligible practices is likely to have been higher than the 32% reported because, among those not responding to the invitation to take part, there will have been some practices which did not use the laboratories for all requests and were thus ineligible.
Although the GPs knew that their PSA requests were being studied they did not appear to alter their use of the PSA test during the prospective period. The results for the 6 months leading up to the prospective period suggest that there was no marked increase in PSA testing as a result of the study itself. There was a possibility that the GPs might not always have distinguished properly between asymptomatic and symptomatic testing, but comparison of the PSA levels between these groups showed that, as expected, the proportion of men with low PSA levels was significantly higher in asymptomatic than symptomatic men. Only 27% of records where a reason was known had the reason recorded using the study stickers; for the remaining 73% of records this relied on extraction of information after the test from patient notes. However, the proportion of reason codes recorded as asymptomatic from the two data sources (stickers and patient notes) was very similar, at 26% and 28%, respectively.
The distribution of levels of PSA show some similarities with results from the European and USA screening trials . In the trials the proportion of men aged 55–69 years with PSA levels of ≥ 4 ng/mL was 8–15% of all men tested per centre. In the present study the corresponding value for asymptomatic testing in men aged 50–69 was 11%, suggesting that men receiving asymptomatic testing may have been a similar population to men in the screening trials. The accuracy of PSA measurements in this study is expected to be comparable among the laboratories, as all of them took part in NEQAS. Although there is a lack of comparability in measurement among different assays at very low and very high PSA levels, the study results are unlikely to have been affected by this, as the PSA levels were analysed in groups, the lowest group being <2 ng/mL and the highest being ≥ 11 ng/mL.
Although the rate of PSA testing is higher in this study than in the previous surveys in the UK [4,5], the rate is likely to be lower than that in the USA, and this difference may be associated with the lower registration rate for prostate cancer in the UK. For example, between 1995 and 1998, 38% of white male populations were tested annually in USA data linking Medicare claims to the Surveillance, Epidemiology and End Results Cancer registry . Comparison of rates of testing between countries is limited by the use of different methods of data collection and measures of rates of testing.
The trends in the rate of PSA testing by social deprivation and ethnicity are likely to reflect differences in the underlying incidence of the disease, in the levels of case finding, awareness of symptoms, use of healthcare and response to screening [9,10]. Further research is needed to explore the reasons for these differences.
The study results illustrate the likely impact on referral of using the PCRMP guidelines, with potentially 14% of asymptomatic and 23% of symptomatic men tested being referred. However, in reality, the referral policies will vary greatly across the country. If for example the GPs in this study had used cut off levels of 3 or 4 ng/mL across all age groups these would lead to 29% or 21% of men, respectively, being referred. The lower the cut off level the greater the proportion of men having biopsies, and the lower the positive predictive value of the biopsy.
The exact proportion of men being screened for prostate cancer in the UK is unknown, but in this study the rate of asymptomatic testing in general practice was at least 1.6%, and certainly higher if private testing had been included. It is uncertain what impact the PCRMP will have on PSA testing in general practice. Given that the overall annual rate of testing is increasing, both in England and Wales, and in Northern Ireland , monitoring the workload and costs in general practice and hospitals is important [12–14], particularly while there remain uncertainties about the effectiveness of screening to reduce mortality . Estimating the workload and costs of referral was beyond the scope of this study.
The study highlights some of the practical difficulties in monitoring PSA testing in England and Wales, as shown by the low response rates from laboratories and general practices. Some of the nonresponse arose from practical difficulties in providing computerized data. In the future, if changes in detection rates of prostate cancer are to be understood, there must be better quality data that are readily available for analysis.
This study was funded by the Policy Research Programme at the Department of Health. The views expressed in this report are those of the authors and not necessarily those of the Department of Health. We thank Mrs Penny Coulson, the study co-ordinator and database manager, Dr Anthony Milford Ward, and Mr Peter White responsible for NEQAS, Dr Karl Meyer of BUPA for the supply of data from this source, Dr Alison Shephard MRCGP for advice in general practice, Mr Richard Birmingham at the Department of Health for the supply of population sizes by general practice, Ms. Angharad Williams in the CSEU for help with data processing, and all GPs involved for providing data.
Acknowledgements of the National Database for primary care groups and trusts as formally requested: The National Database for Primary Care Groups and Trusts is a product of the National Primary Care Research and Development Centre at the University of Manchester. It was devised by Dr Deborah Baker, who is the database director. The database was constructed by Justin Hayes at the Regional Research Laboratory, School of Geography, University of Manchester (Director: Dr Robert Barr); SEE IT consultancy designed and built the map interface. We are grateful to Andrew Wagner, Mark Hann and David Reeves (NPCRDC) for their considerable hard work in cleaning and validating the data sets. Andrew Wagner is the database manager (firstname.lastname@example.org).
None declared. Source of funding: Department of Health.
Dr J Fyffe Mr C Perrett, Queen Elizabeth II Hospital, Welwyn Garden City, Herts, Dr P Raggatt Dr G A Maguire, Addenbrookes Hospital, Cambridge, Dr Everitt Ms. K Roche, Basildon Hospital, Basildon, Essex, Dr M. Browne Dr W De Ste Croix, B Patel, The Royal London Hospital, London, Dr P O Collinson Mr S Moody Dr F Boa, St. George's Hospital, London, Dr P Foley, Whipps Cross Hospital, Leytonstone, London, Dr I Watson Mr E Jackson, Southport & Ormskirk Hospital, Southport, Dr Gl Burrows Mr G Tonks, Stepping Hill Hospital, Stockport, Dr A Davis Mr A Rawsthorn, Warrington Hospital, Warrington, Cheshire, Dr D Hirst Dr D Robertshaw, Bradford Royal Infirmary, Bradford, Dr M Sinclair Dr A Waise, York District Hospital, York, Dr Waise Mr D Parrington, Friarage Hospital, North Allerton, Nth Yorkshire, Dr J R Quiney Mrs C. Chowne, St. Richards Hospital, Chichester, Mr R Walker Dr D Rowe, Southampton General Hospital, Southampton, Dr J Barron Dr J. Haite Mr S. Bunnage, St. Helier Hospital, Carshalton, Surrey, Dr S Fleming Mr B Thomas, Royal Cornwall Hospital, Truro, Cornwall, Dr D James Mr B Jones, Taunton & Somerset Hospital, Taunton, Somerset, Dr M Salzmann Mrs D Perry Mr R Owens, Royal Devon & Exeter Hospital (Wonford), Exeter, Devon, Dr G Griffiths Mr R Osgerby, County Hospital, Lincoln, Dr P Masters Mr T Wilson, Chesterfield Royal Hosp, Chesterfield, Derbyshire, Dr J Wardell Dr R Stott Mr M Slokan, Doncaster Royal Infirmary, Doncaster, Dr C P Williams Mr G K Davies, Wrexham Maelor Hospital NHS Trust. Wrexham, Dr D A Hullin Mr P Evans, Royal Glamorgan Hospital, Rhondda-Cynon-Taff, Wales, Dr M Giles Mr E R M Jones Ms. G Edwards, Glad Clwyd Hospital, Rhyl, Denbighshire, Dr Haboubi Dr K Powell, West-Wales General Hospital, Carmarthen, Dr S Ramachandran Mr B Downing Ms. K Taylor, Good Hope Hospital, Sutton Coldfield, W. Midlands, Dr T Reynolds, Queens Hospital, Burton Upon Trent, Staffordshire, Mr G C Mascal, Worcestershire Acute Hosp. NHS Trust, Kidderminster, UK BUPA laboratory collated by Dr Karl Meyer, Leeds, UK
The method used to adjust rate of testing for ‘reason for test not known’ and to the social deprivation distribution of England.
(a) Adjustment for reason for test not known.
Using data in Table 2, the proportions of the total number of tests by reason were calculated within each age group, excluding the group with reason not known. The total number of tests with reason not known were then split according to these proportions and the relevant number of tests added to each of the known reason groups.
(b) Adjustment to the social distribution of England.
By comparison with the distribution of Townsend social deprivation scores for all Primary Care Groups in England, the distribution of scores for the Groups covered by study practices represents a more affluent subset. The distribution of Townsend scores for study practices was divided into quintiles to define five score ranges and the proportion of the all-England scores falling within each of the five groups was calculated. The relative size of the all-England proportion vs the study proportion was used to derive a correction factor for each quintile. For example, 20.33% of study area scores fell within the first quintile, but only 15.92% of scores for the whole of England fall within this group, resulting in a correction factor of 0.783. The correction factors (Table 4) were applied to the rates of testing by reason within each group of study practices defined by the quintiles.
|Reason||Quintiles of Townsend deprivation score|
|−4.5 to −2.81||−2.80 to −1.93||−1.92 to −1.06||−1.05 to 0.52||−0.53 to 14.7|
|% (no. of tests)|
|Asymptomatic||1.1 (1054)||0.8 (742)||0.8 (735)||0.7 (608)||0.4 (272)|
|Symptomatic||1.3 (1185)||1.5 (1343)||1.0 (902)||1.1 (964)||0.8 (560)|
|Re-test||0.6 (591)||0.7 (636)||0.4 (397)||0.4 (328)||0.3 (187)|
|Monitoring cancer||0.5 (499)||0.6 (529)||0.3 (296)||0.3 (286)||0.1 (100)|
|Not known||1.0 (917)||0.7 (621)||0.4 (361)||0.6 (518)||0.7 (511)|
|Total||4.5 (4246)||4.4 (3871)||3.0 (2691)||3.0 (2704)||2.2 (1630)|
|Male population aged 45–84 years||93 451||88 546||88 599||89 799||74 479|