Impact of PSA testing and prostatic biopsy on cancer incidence and mortality: comparative study between the Republic of Ireland and Northern Ireland
- First Online:
- Cite this article as:
- Carsin, A., Drummond, F.J., Black, A. et al. Cancer Causes Control (2010) 21: 1523. doi:10.1007/s10552-010-9581-y
- 111 Views
To investigate the impact of different PSA testing policies and health-care systems on prostate cancer incidence and mortality in two countries with similar populations, the Republic of Ireland (RoI) and Northern Ireland (NI).
Population-level data on PSA tests, prostate biopsies and prostate cancer cases 1993–2005 and prostate cancer deaths 1979–2006 were compiled. Annual percentage change (APC) was estimated by joinpoint regression.
Prostate cancer rates were similar in both areas in 1994 but increased rapidly in RoI compared to NI. The PSA testing rate increased sharply in RoI (APC = +23.3%), and to a lesser degree in NI (APC = +9.7%) to reach 412 and 177 tests per 1,000 men in 2004, respectively. Prostatic biopsy rates rose in both countries, but were twofold higher in RoI. Cancer incidence rates rose significantly, mirroring biopsy trends, in both countries reaching 440 per 100,000 men in RoI in 2004 compared to 294 in NI. Median age at diagnosis was lower in RoI (71 years) compared to NI (73 years) (p < 0.01) and decreased significantly over time in both countries. Mortality rates declined from 1995 in both countries (APC = −1.5% in RoI, −1.3% in NI) at a time when PSA testing was not widespread.
Prostatic biopsy rates, rather than PSA testing per se, were the main driver of prostate cancer incidence. Because mortality decreases started before screening became widespread in RoI, and mortality remained low in NI, PSA testing is unlikely to be the explanation for declining mortality.
KeywordsProstate cancerTrendsPSA testsIncidenceMortality
Annual percentage change
National Cancer Registry Ireland
Northern Ireland Cancer Registry
Republic of Ireland
Prostate cancer mortality rates have declined in the United States and some European countries [1–4]. Whether the widespread use of PSA testing in the US and other countries has influenced mortality at the population level is unknown and heavily debated [3, 5–8]. The recently published results of two large trials [9, 10] have not resolved this issue but have, instead, contributed further to the debate. Although no reduction in prostate cancer mortality was observed in the PLCO Cancer Screening Trial in the United States , results from the European Randomized Study of Screening for Prostate Cancer (ERSPC) showed that screening via PSA testing resulted in a 20% decrease in prostate cancer mortality after 9 years . However, the ERSPC results do not necessarily imply that the observed decline in prostate cancer mortality starting in the early nineties is due to screening and early detection, and further data are needed to clarify this issue.
Comparisons between countries with different policies and practices concerning prostate cancer detection have an important role to play in explaining incidence and mortality trends. Since temporal trends in prostate cancer are notoriously difficult to interpret , such comparative studies should ideally also include population-level data on PSA testing and prostatic biopsies. Most previous studies have been limited by a lack of such data [1, 2, 11, 12].
The Republic of Ireland (RoI) was estimated to have the highest prostate cancer incidence rate in Europe in 2006 . PSA testing is used extensively in primary care in the RoI, but there are wide variations in practice, likely to be due in part to a lack of national guidelines . The RoI has a mixed public–private health care system, with the majority of visits to primary care incurring a charge to the patient. By contrast, in Northern Ireland (NI), primary and secondary health-care is delivered by the publicly funded National Health Service (NHS) and is free at the point of delivery. The NHS does not recommend the use of PSA testing for prostate cancer screening in primary care , and men access urologists via the public system with a small amount of private practice. Despite this, there is some evidence of PSA being used for screening in NI . The contrast in the health-care systems and PSA testing policies between these countries with similar populations makes them an ideal setting in which to investigate factors influencing prostate cancer detection and mortality rates. It was against this background that we conducted a population-based comparison of time trends in PSA testing, prostatic biopsies and prostate cancer incidence and mortality in the RoI and NI.
Materials and methods
Prostate cancer incidence
Anonymized data on invasive prostate cancers (ICD-O2: C61) were obtained from the National Cancer Registry Ireland (NCRI) (www.ncri.i.e.) and the Northern Ireland Cancer Registry (NICR) (www.qub.ac.uk/nicr), for cases diagnosed from the first year of registration (1993 in NI and 1994 in the RoI) until 2005. Cases were excluded where diagnosis was made by death certificate only (representing 2% and 1% of cases in the RoI and NI, respectively). Details on age and year of diagnosis were abstracted. In NI, information on tumor grade was extracted from histopathology reports between 1996 and 2005. In the RoI, these data were routinely recorded as part of the cancer registration process by the trained tumor registration officers using the WHO grading . When information on Gleason score (GS) only was available in the medical records, this had been translated previously by the registry into WHO grading as follows: GS < 5 = grade 1; GS 5–7 = grade 2; and GS > 7=grade 3.
Data on the number of PSA tests by year were extracted from databases compiled by the two registries. Information on PSA tests performed in NI since 1994 is routinely collected by the NICR from all laboratories which measure PSA (n = 10). Tests done in the period 1994–2005 were tabulated by year of, and age at, testing. In the RoI, a special data collection exercise was conducted to obtain PSA data and is described elsewhere . Briefly, data on all tests conducted during 1994–2005 were sought from the 36 laboratories which measure PSA. Individual-level data for all tests performed in each year were sought, but not all laboratories could provide this. To estimate the annual numbers of tests nationally, we allocated a weight to each laboratory by year based on the number of PSA tests they reported performing in the first quarter of 2006 and the year they started measuring PSA . Comparing these numbers with the individual-level data provided, we estimated that we collected information on 58% of the total number of tests conducted during 1994–2005.
Data on PSA tests were linked to the NCRI database by name, date of birth and address (where available). A similar linkage was performed in NI and described elsewhere . Any tests performed after the date of diagnosis of cancer were excluded.
Prostatic biopsy data
In the RoI, numbers of prostatic biopsies, by year and age-group, were obtained from the Hospital In-Patient Enquiry System (HIPE) which records all discharges from all public hospitals. Prostate biopsies were defined as any discharge record with an ICD9 procedure code in the range 60.11–60.15. Anonymized data on claims for all biopsies performed in private hospitals, by year and age group, were obtained from the two health insurance companies, in operation in the RoI during the study period, VHI Healthcare and BUPA. Since these data were only available from 1996 onwards, the analysis of prostatic biopsies in the RoI was limited to the period 1996–2004. In NI, information was obtained for the period 1999–2004 from the Directorate of Information Services, which records procedure codes from all hospital discharges in NI. Data were not available for years prior to 1999. Only needle biopsies of the prostate were included; other procedures (e.g., transurethral resection of the prostate) were excluded. Only total counts could be provided (i.e., these data could not be broken down by age).
Prostate cancer mortality data
Mortality data were extracted from World Health Organization mortality database for the period 1979–2006. For the RoI, prostate cancer deaths were defined as those coded to ICD9 185. In NI, ICD9 was used for the period 1979–2000 and was replaced by ICD10 (code C61) from 2001 onward. The change in classification did not alter prostate cancer rates in NI .
All rates were expressed in relation to the male population aged ≥50 years, using mid-year population estimates obtained from the RoI Central Statistics Office  and the Northern Ireland Statistics and Research Agency. Incidence, mortality and PSA testing rates were standardized using the European Standard Population restricted to age ≥50 years. The prostatic biopsy data for NI could not be age-standardized, so only crude rates are presented for NI. In the RoI, biopsy rates were standardized to the NI age-structure to permit comparison between the two countries. Differences in rates between the RoI and NI were assessed using a χ2 test. Annual percentage changes (APC) in incidence, mortality, PSA testing and prostatic biopsies, and the points in time when trends changed, were estimated by fitting joinpoint regression models , using joinpoint software (3.1) . This method determines the number of significant joinpoints (time at which there is a change in slope) by performing permutation tests and fits a log-linear model to each segment. Trends in incidence by grade were similarly computed, with grades 1–2 combined and classified as “low/medium-grade” and grade 3 as “high-grade” to simplify analysis.
The Cuzick non-parametric test of trend was used to investigate whether median PSA level within 6 months prior to a prostate cancer diagnosis, and median age at diagnosis, changed significantly over the period investigated. Difference in median age at diagnosis between the RoI and NI was assessed using the Wilcoxon rank-sum test.
To further quantify changes in prostate cancer incidence, we estimated the number of “extra cases” in each country, assuming that in the absence of PSA testing, the true underlying incidence would not have changed over time. We calculated the number of expected cases in each year by applying the 1994 5-year age-specific rates to the corresponding population counts for that year. Excess cases were estimated by the difference between observed and expected cases.
Prostate cancer incidence rates
Rates over time for prostate cancer incidence and mortality, PSA testing and prostatic biopsies, with annual percentage changes (APC)
On the basis of incidence rates in 1994, there were 5,938 and 763 extra prostate cancer cases diagnosed, respectively, in the RoI and in NI between 1995 and 2005. This is 42% (i.e., 5938/[19844-5938]) and 12% (i.e., 763/[7388-763]) more, respectively, than if 1994 rates had persisted until 2005.
The median age at cancer diagnosis was significantly lower in the RoI (71 years) compared to NI (73 years) (p < 0.01) and decreased significantly over time in both countries (RoI: 1994, 74 years; 2005, 68 years (p-trend < 0.01); NI: 1994, 74 years; 2005, 70 years (p-trend < 0.01)).
PSA level prior to a cancer diagnosis
Median PSA level in tests within 6 months prior to cancer diagnosis
Prostate cancer mortality
This paper describes the impact of differing rates of prostate cancer investigation on the rate of this disease in similar populations. We demonstrate the effect of screening on the average age at diagnosis, as more younger men are diagnosed and at a lower PSA level. We also quantify the excess number of cases diagnosed as a result of PSA testing.
Our results suggest that the decline in prostate cancer mortality in the RoI and NI is not associated with opportunistic screening via PSA testing: despite widely different levels of investigation and treatment, and the increase in low/medium grade tumors, decreases in mortality rates were equivalent in both countries. These decreases pre-date widespread PSA testing and occurred long before an effect would be expected given the estimated lead time of 5–12 years [8, 24, 25]. Our findings are consistent with a study performed in the United Kingdom by Hussain et al. . Although their observation relied on PSA testing rates, and not biopsy rates, they observed a decline in mortality which could not be attributable to PSA testing.
Radical prostatectomy has been found to reduce prostate cancer mortality if done when the cancer is still organ confined . But in Ireland, in 1996, fewer than 5% of cancer patients underwent this operation [27, 28]. There is growing evidence that hormone deprivation therapy alone or as an adjunct to radical surgery or radiotherapy can improve local control of localized and locally extended prostate cancer and increases survival of prostate cancer patients [29, 30]. No information was available on trends in hormonal treatment prior to 1996, but in that year surveys conducted by national cancer registries in both countries showed that 40% of prostate cancer patients in the RoI  and 68% in NI received this form of treatment . Therefore, the introduction of pharmaceutical hormonal therapy [2, 32] could partly explain the fall in mortality rates. Increased use of hormonal therapy may have impacted on prostate cancer mortality [11, 33] by delaying death in patients with advanced disease  who may then die of other causes.
Recent surveys in the UK, NI and the RoI suggest that greater disease awareness among the public and health professionals is another factor that may have contributed to decreasing mortality [35–37]. For instance, greater awareness of prostate cancer may have led to more systematic PSA testing in men presenting with urinary symptoms or to increasing demand for PSA testing by men with a family history of prostate cancer. More intensive monitoring of patients after diagnosis could also have played a role.
Differences in death certificate coding practices may explain the slightly higher mortality rate in the RoI around 1976. In the 1980s and after, attribution bias  may have been a factor in the higher mortality in the RoI: the higher prostate cancer incidence in the RoI might have caused more deaths of unknown cause to be attributed to a previously diagnosed prostate cancer. To test this hypothesis, we compared mortality rates in the RoI and NI in 5-year age bands. Mortality rates were similar in the two countries among the younger age groups (50–69 years) whereas the rate was higher in the RoI in every band from 70 onwards, and this difference increased with increasing age. The greater difference in peak mortality rates, compared to those at the start and the end of the observation period, also supports the likelihood that attribution bias can explain some of the differences in mortality between the RoI and NI.
The rate of PSA testing, however, cannot fully explain the observed difference in prostate cancer incidence rates between the two countries. Between 1994 and 2000, when rates of testing were similar, cancer incidence rates were consistently higher in the RoI. In NI, during 1994–1999, incidence remained stable despite a 10% annual increase in the rate of PSA testing. On the other hand, the trends in incidence and biopsy rates were concurrent with incidence rates starting to rise as biopsy rates rose from 2000 onwards. In the RoI, age-specific trends in biopsy closely mirrored the time trends in prostate cancer incidence.
In addition to higher PSA testing rates in RoI than in NI, there was also evidence that the threshold for biopsy was lower as suggested by the observed lower median PSA level at cancer diagnosis. The results of our research among GPs and urologists in NI and the RoI also point to this conclusion. This found that, during the study period, in NI men were unlikely to be referred for biopsy unless their PSA level was greater than 10 ng/ml , whereas in the RoI the majority of GPs would refer a man at a PSA level of 7 ng/ml . The younger age at cancer diagnosis of men in the RoI compared with men in NI also supports the hypothesis that men were being investigated more intensively in the RoI for prostate cancer. In this respect, our findings clearly demonstrate that the biopsy rate, and not the PSA testing rate per se, is the main driver of increasing prostate cancer incidence. Therefore, we suggest that information on PSA testing alone [11, 39, 40] is not sufficient to study the impact of prostate cancer screening activity; reliable information on biopsy rates is essential.
The rapid increase in biopsy rates in NI from 2000 onwards deserves some comment. We are not aware of any change in biopsy referral guidelines, or guidance, around this time. One contributing factor is likely to be the rise in numbers of urologists—from 10 in 1999 to 15 in 2005—thus increasing capacity for biopsies. It is also likely that urologists’ views about the “appropriate” PSA threshold for referral have changed over time; while in the past it was conventional to consider a raised PSA as 10 ng/ml or above, nowadays levels of 3–4 ng/ml are considered raised.
Increasing PSA testing, higher biopsy rates, and the observed decrease in PSA threshold for biopsy at cancer diagnosis  lead to the possibility of over-diagnosis and over-treatment. In the RoI, for example, 5,938 men were diagnosed with prostate cancer in 1995–2005 who would not have been diagnosed had PSA testing not been introduced. This corresponds to a relative increase of 30% in the number of cases diagnosed, a value which falls between the excesses found in the PLCO (22% after 7 years)  and the ERSPC trials (70% after 9 years) . Most of these men would probably not have died from prostate cancer, and some will have undergone unnecessary treatment with life-changing side effects .
Although GPs may find it difficult to follow guidelines on PSA testing while under pressure from patient requests, referral for biopsy is under the GP’s control to a greater extent. This study suggests that guidelines for referral and biopsy would be more effective than guidelines on PSA testing in targeting investigation and treatment to those men most likely to benefit. Further evaluation of PSA, particularly with regard to management of equivocal results and diagnostic follow-up, is important to ensure that the potential benefits are maximized and the associated harms are minimized.
Limitations of the study
The main limitation of our study was that it was ecological. We could not obtain exhaustive data on PSA testing in the RoI. It is, however, likely to be representative: from our data, we estimated that in 2004, 14% of men in the RoI and 6% in NI had had a PSA test. These values are very close to those from the Eurobarometer survey  which reported that 15% of men in the RoI and 8% of men in the UK had had a PSA test in 2006.
Prostate cancer mortality in both NI and the RoI began to decrease before PSA testing became widespread. Moreover, the increase in incidence in both countries in the past two decades and the persistently higher rates in the RoI compared to NI were more closely related to differences in prostatic biopsy rates following PSA testing in men than to the level of PSA testing. These observations suggest that prostatic biopsy rates, rather than PSA testing “per se”, are the main driver of prostate cancer incidence. Consequently, we suggest that prostatic biopsy rates may better reflect screening activities in a population than PSA testing rates. On the basis of this, we also suggest that guidelines on the threshold at which men should be referred for biopsy following PSA testing are urgently required. As regards the observed falls in mortality, it is possible that these are linked with the widespread adoption of modern pharmaceutical hormone therapy as a standard therapy for prostate cancer in the early 1990s.
There is an imperative to continue research in this area. In these populations alone, more than 6,700 extra cases of prostate cancer (85 per 100000 men aged 50+, annually) were diagnosed and treated with no apparent effect on deaths. Similar excesses are likely to be evident in other countries. Since PSA testing is undoubtedly here to stay, the challenge for the research and clinical communities is to find ways to maximize the benefits and, most importantly, minimize harms both for individual men and for the population as a whole.
We are grateful to the staff of the Northern Ireland Cancer Registry and the National Cancer Registry Ireland who collected and processed the cancer registration data used in this paper. In particular, we thank Mr Colin Fox, Ms Wendy Hamill and Mr Richard Middleton for their help in providing the data from NI and Christine Buicke for reviewing death certificates in the RoI. We are also grateful to colleagues at HIPE, VHI Healthcare and BUPA for providing biopsy data, and staff in hospitals and laboratories for providing PSA data.
Competing interest statement
All authors declare that the answer to the questions on your competing interest form are all no and therefore have nothing to declare.
Details of contributors
AEC did the statistical analyses. FD collected and provided data on biopsies and PSA tests in the RoI. AB provided the data on PSA tests in NI. PL helped with prostate cancer grade and stage data collection in NI. DC provided the biopsy data in NI. FD, AB, LS, LM, PA, MB, HC and AG took part in the study design. AEC, FD, LS, PA, LE, HC, AG wrote and revised the article. All authors were involved in revisions and approved the final version. AEC is the guarantor for the study.
The Northern Ireland Cancer Registry is funded by the Public Health Agency Northern Ireland, and the National Cancer Registry of Ireland is funded by the Department of Health and Children (Republic of Ireland). Some aspects of the data collection for this study were funded by a grant from the Northern Ireland Research & Development Office and the Health Research Board (Dublin) [grant number NS/2004/20]. The study funders had no role in the study design, the collection, analysis and interpretation of the data. All authors were independent from the funding source.