Mathieu Boniol, International Prevention Research Institute, 95 cours Lafayette, 69006 Lyon, France. e-mail: firstname.lastname@example.org
Study Type – Therapy (data synthesis)
Level of Evidence 2b
What's known on the subject? and What does the study add?
The efficacy of prostate cancer screening using PSA testing is still being debated, with conflicting results in randomized trials.
The study shows that, even using the hypothesis most favourable to prostate cancer screening with PSA, the net number of years of life does not favour screening.
• To evaluate the impact of the implementation a prostate-specific antigen (PSA) screening programme using the European Randomized Study of Screening for Prostate Cancer (ERSPC) results and taking into account the impact of prostate biopsy and over-treatment on mortality.
MATERIALS AND METHODS
• We used a model based on the number of years of life gained and lost owing to screening, using data reported in the ERSPC.
• We conducted a critical evaluation of the ERSPC results and of the Swedish arm of the study.
• Accounting for biopsy-specific mortality and for over-treatment, the balance of number of years of life was negative in the ERSPC study, with an estimated loss of 3.6 years of life per avoided death.
• The number of years of life becomes positive (real gain) only when fewer than 666 screened individuals are required to avoid one death.
• We found that in the Swedish arm of the ERSPC there was a biopsy rate of 40% compared with 27% in the ERSPC overall. The over-treatment rate was also greater with 4.1% compared with 3.4% overall.
• For the last 20 years, there has been a marked difference in prostate cancer-specific mortality between Sweden and the rest of Europe: in 2005, for the age group 65–74 the rate was 140 per 100 000 person years in Sweden and ∼80 per 100 000 for the rest of Europe.
• Overall, PSA testing in Europe is associated with a loss in years of life and should thus not be recommended.
European Randomized Study of Screening for Prostate Cancer
Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial
Institut National de la Statistique et des Etudes Economiques.
The validity of mass screening for prostate cancer is still a matter of discussion because the conclusions of various randomized studies are conflicting [1–4]. In 2009, the multicentric European Randomized Study of Screening for Prostate Cancer (ERSPC) reported a significant decrease in prostate cancer mortality (risk ratio [RR] 0.80; 95% CI 0.65–0.98). In the meantime, the US Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO) failed to show an impact on cancer-specific mortality (RR = 1.13; 95% CI 0.75–1.70). Djulbegovic et al. , in their meta-analysis of five randomized trials, concluded that there was no significant reduction in prostate cancer-specific mortality associated with prostate cancer screening by PSA testing. The authors further concluded that the Swedish arm of the ERSPC  was an outlier as compared with other studies.
Recently, both the ERSPC and the PLCO reported updated data with follow-up of 11 years and 13 years, respectively [6,7]. These data showed the same discrepancies: the RR was 0.79 (95% CI 0.68–0.91) for the ERSPC and 1.09 (95% CI 0.87–1.36) in the PLCO. Criticisms of the methodology of ERSPC still applied .
In the absence of a clear evaluation of the value of prostate cancer screening with PSA testing, the current recommendation is limited to a shared decision between the patient and the urologist [8,9]. The US Preventive Services Task Force recently recommended against PSA-based screening for prostate cancer , but PSA testing remains widely promoted and used outside of the usual requirements for evidence-based medicine.
To assess the benefit of mass screening, a decrease in cancer-specific mortality is a necessary condition but is not sufficient . The global health impact should be evaluated by including drawbacks that result from the implementation of screening such as over-treatment, side effects, and the general impact on quality of life. Prostate cancer screening is associated with a large proportion of over-diagnosis  and over-treatment, and it can therefore lead to unnecessary treatments which can, in turn, have serious side effects. The use of PSA screening also increases the rate of prostate biopsy in the targeted population and thereby increases the side effects of biopsy. A high rate of biopsy has been reported in all randomized trials.
The randomized trials currently available are not sufficiently powered to detect a significant impact on all-cause mortality , and do not provide assessments of over-treatment and over-diagnosis; therefore, we made use of modelling approach to evaluate the overall impact of prostate cancer screening on all-cause mortality.
The objective of the present study was to estimate the gains and losses associated with a screening intervention, based on the hypothesis most favourable to screening (the recent update of the ERSPC study), in the population in France of men aged 55–64 years old.
MATERIAL AND METHODS
The modelling approach to calculating the number of years of life lost and gained by screening is a good method for evaluating the overall impact of a prostate cancer screening programme. This calculation is particularly relevant for prostate cancer, as testing is generally conducted on an older population. In such a population, screening a patient at 50 years old would have a potential impact several years later while side effects could come earlier. This population is also more fragile and likely to be more sensitive to side effects. Our method was based, therefore, on the calculation of the balance of years of life gained and lost from prostate cancer screening. This calculation involved pooling information on the efficacy of screening, the biopsy rate in the ERSPC, life expectancy for men in the age group targeted for screening, years of life gained per prostate cancer death avoided, the mortality rate associated with prostate cancer treatment, and the mortality rate after prostate biopsy.
We extracted the following data on PSA screening from the update of the ERSPC with 11 years of follow-up: the number of prostate cancers detected, and the number of patients invited for screening for one prostate death avoided. We found that one prostate cancer death was avoided for 37 prostate cancers treated and 1055 individual invited for screening.
Biopsy rate information was extracted from the original publication [3,6] as the number of biopsy procedures per individual in the screening arm. The biopsy rate was 27% in the ERSPC.
Each prostate death avoided contributed to 9.3 years of life gained, based on an estimation from the Surveillance Epidemiology and End Results statistics on the number of years of life lost per person dying from prostate cancer in 2006 . Because any death attributable to prostate cancer treatment, or as a consequence of the biopsy, would occur rapidly in the follow-up, we considered the life expectancy at the age of 65 years old when calculating years of life lost. According to the Institut National de la Statistique et des Etudes Economiques (INSEE) statistics, at age 65 life expectancy was 17.1 years in 2003 . In other words, a death avoided from prostate cancer would contribute 9.3 years of life gained while a death consequent to a biopsy would contribute 17.1 years of life lost.
Mortality attributable to prostate cancer treatment has been evaluated in various studies . We retained the rate of five deaths per 1000 prostate cancers treated .
The side effects of prostate biopsy are usually limited to infections, further morbidity  and, rarely, to deaths resulting from septic shock; however, we identified only one study  that produced population-based statistics on deaths after prostate biopsy: the specific rate was 0.2% for men aged < 60 years old, 0.3% for men aged 61–65 years old and ∼ 0.4% for men aged 66–70 years old. We therefore used this value of two deaths per 1000 biopsies.
We applied the calculation of years of life gained and lost by applying these figures from prostate cancer screening to the French population. From INSEE statistics in 2009 , 4097 560 men aged 55–64 years old would be targeted for a population-based screening.
We completed our modelling of the years of life balance using one-way sensitivity analysis in which one factor varied while the other remained unchanged. We did this to find out which conditions had to be met to have the balance set to null, i.e. no gain and no loss in years of life. Similarly, we conducted two-way sensitivity analysis in which two factors varied at the same time: first, mortality associated with prostatic biopsy and mortality associated with prostate cancer treatment; second, mortality associated with prostatic biopsy and with biopsy rate.
Because, the various trials to date were conducted on populations with different background mortality rates, our aim was to investigate whether the different results published so far could be related to the mortality rate; therefore, we extracted data from the WHO mortality database. We used the International Classification of Diseases-10 definition with code C61. We calculated the mortality rate for the year 2005 for the age groups 55–64 years and 65–74 years.
Table 1[14,16] shows the calculation of the balance in years of life with data from the ERSPC study, i.e. with one prostate cancer death avoided for 37 prostate cancers diagnosed and 1055 individuals invited for screening. Deaths associated with treatment after prostate cancer diagnosis represent 3.2 years of life lost. Prostate biopsy overall represents 9.7 years of life lost. The balance accounting for the 9.3 years of life gained by one prostate death being avoided shows a mean loss of 3.6 years of life.
Table 1. Calculation of years of life gained and lost from prostate cancer screening
When applied to the population of men in France aged 55–65 years old, the number of prostate cancer deaths avoided was estimated to be 3884 deaths, the number of deaths generated by the screening would be 2927 deaths, 719 from prostate cancer treatment and 2209 from biopsy mortality. This would mean a balance of 957 deaths avoided; however, when expressed in number of years of life, the balance would be negative, with 13 937 years of life lost (36 121 years gained by avoiding prostate cancer deaths, 50 057 years lost by prostate treatment and deaths after biopsy).
In a sensitivity analysis, the overall balance of years of life is mainly influenced by the biopsy rate (Fig. 1) and by the mortality associated with prostate biopsy (Fig. 2). The number of years of life would become null, i.e. not in favour of or against prostate cancer screening, if the biopsy rate was 17% or if the prostate biopsy mortality rate was decreased to 1.26 deaths per 1000 biopsies. Other factors can only affect the balance of years of life to the null for unrealistic values such as a mean gain of >13 years for each prostate death avoided, or a mean life expectancy at 65 years of <12 years. Even if we assumed no deaths attributed to prostate cancer treatment, the balance for one death avoided would still be negative with a mean loss of 0.4 years of life. In the two-way analysis, making both mortality from prostate biopsy and from treatment vary, the impact of treatment on the balance of years of life becomes important only if mortality from prostate biopsy is reduced (Fig. 3). In the present sensitivity analysis, if mortality from prostatic biopsy was reduced to 0.1%, a further decrease of mortality from treatment from 0.5% to 0.25% would shift the gain of almost 3 years of life to >4 years of life.
In a further two-way sensitivity analysis, mortality from prostate biopsy and the biopsy rate varied together (Fig. 4). If the mortality from prostate biopsy was halved to 0.1% but at a cost of an increase in the rate of biopsy to 40%, the net balance in years of life would still be above the black curve, i.e. in disfavour of screening with a small loss in years of life.
We calculated the balance in years of life gained and lost for various numbers of patients needed to invite for screening for the ERSPC. The relationship between gain and loss in years of life after screening is affected by the number needed to screen: the balance becomes positive only if the number needed to invite for screening is <666.
Figure 5 shows the prostate cancer-specific mortality rate in 2005 for countries participating in the ERSPC and selected countries. Sweden has the highest mortality rate in the age group covered by the screening, a rate twice that of the USA for the age group 65–74 years.
We evaluated the balance of the number of years of life gained and the number lost after prostate cancer screening and considered the impact of over-treatment and the side effects of prostate biopsy. The study was limited to the statistics from the ERSPC study , which was the only trial showing a significant decrease in prostate cancer-specific mortality. Hence, under this best hypothesis condition for screening efficacy, it appears that screening would generate more harm (in terms of lost years of life) than good. The balance between side effects and efficacy becomes in favour of screening only with highly effective screening for which a prostate death would be avoided for <666 individuals invited for screening.
The present data are driven mainly by the biopsy rate and by the mortality rate after prostate biopsy. Other factors only play a minor role in the estimation of the impact of screening. The mean life expectancy included in our model was close to that of other European countries or in the USA , with life expectancy at 65 years of ∼17 years.
From comparison of prostate cancer mortality in various countries, it appeared that Sweden had and still has a particularly high rate of prostate cancer-specific mortality. This could be an important factor for the potential efficacy of PSA screening which requires a minimum prevalence of the disease to be really efficient as a mass screening diagnostic tool. The difference between Sweden and other countries in term of prevalence of disease might even be higher than that depicted by the mortality rate. Such a difference could result in the particular result observed in the Swedish arm of the ERSPC .
Several elements could explain the difference in screening outcomes between the Swedish arm of ERSPC, which showed a benefit for those participating in screening, and the remaining ERSPC centres, which resulted in a negative evaluation. Firstly, there is clearly a higher prostate cancer-specific mortality rate in Sweden than in other European countries, hence the pool of potential individuals at risk, and who could benefit from earlier detection, is higher. Secondly, the biopsy rate in the Swedish arm of the ERSPC was higher than in other countries, with 40% of eligible men screened in the Swedish arm ever having a biopsy compared with 27% overall in the ERSPC. This considerably higher biopsy rate (nearly half of all men screened) could be explained by the lower PSA threshold used in the Göteborg centre . The decrease in PSA threshold values was associated with a larger degree of over-treatment. The role of PSA is questionable when nearly half the men will finally have had a biopsy. Furthermore, our two-way sensitivity analysis showed that even if mortality from prostate biopsy was reduced by 50% (i.e. to 10 per 10 000 biopsy) a biopsy rate of 40% would make the net balance in years of life still in disfavour of PSA screening.
The low specificity of PSA testing is counterbalanced by a very high rate of biopsy in the population. Almost half of the men are diagnosed with a high PSA level which is followed by a prostate biopsy. The PSA threshold might indeed be crucial to the rate of biopsy when comparing with other studies; in PLCO, despite the pre-existence of PSA screening with a threshold of 4 ng/mL, the biopsy rate was 12.7% .
The role of the mortality rate associated with prostate biopsy seems surprising at first glance as the medical procedure seems relatively benign. The mortality after biopsy data were extracted from one single study conducted between 1989 and 2000 in Canada . It is, to our knowledge, the only specific estimation of the death rate after prostate biopsy. Because of the important role of this mortality rate, this statistic of mortality associated with prostate biopsy should be viewed with caution as it might not reflect the best current practice. Another study conducted between 1996 and 2005, although not specifically designed to investigate this, reported a mortality rate associated with biopsy of 0.9% . In the model used in the present study, the use of this higher value would be even in more in disfavour of prostate cancer screening. Further studies should focus on the exact impact of prostate biopsy in terms of sceptic shock but also on quality of life.
The present study focused on overall deaths associated with prostate screening, in the best current hypothesis based on results from the ERSPC study. But this evaluation does not account for factors external to overall death such the evaluation of quality of life  which would also be not in favour of prostate cancer screening. We also did not account for cost evaluation . With regard to the cost aspect, economic evaluation usually implies that costs and benefits occurring at different points in time do not have the same value; therefore a discount should be applied to events occurring later in time . If we have applied a discount rate of 9%, which would correspond to a ratio of 1:4 at 15 years of difference (in other words, an individual would give the same weight for a loss of 1 year of life at 65 years or the loss of 4 years of life at 80 years), in such a case, the evaluation of the ERSPC study would become even lower to a balance of a mean loss of 8 years of life.
The update of the ERSPC  showed an improvement in efficacy with longer follow-up , with a lower number of men needed to invite per avoided death, from 1410 in 2009 to 1055 in 2012, and we could hypothesize that a prolonged follow-up would be associated with further improvement; however, it should be kept in mind that the biopsy rate would also continue to increase in an ageing population, a population with an even higher risk associated with biopsy. The negative effects described in the present paper would then at least run in parallel and more likely increase, with a prolonged follow-up.
From the present evaluation under the best case scenario of the impact of prostate cancer screening, the implementation of PSA testing in the general population cannot be recommended as public health policy. Unless the mortality rate associated with prostate biopsy can be decreased, screening for prostate cancer with PSA should be discouraged.