The impact of design and performance in PSA screening: differences between ERSPC centers Eveline A.M. Heijnsdijk1, Jan Adolfsson2, Anssi Auvinen3, Monique J. Roobol4, Jonas Hugosson5, Harry J. de Koning1.
1. Department of Public Health, Erasmus Medical Center, Rotterdam, the Netherlands
2. Department of Clinical Science, Intervention and Technology, Karolinska Institutet, Stockholm, Sweden
3. School of Health Sciences, University of Tampere, Tampere, Finland
4. Department of Urology, Erasmus Medical Center, Rotterdam, the Netherlands 5. Department of Urology, Sahlgrenska University Hospital, Göteborg, Sweden
Word count: Abstract: 200 Text: 986
Corresponding author: Eveline Heijnsdijk, PhD Department of Public Health Erasmus MC
PO BXO 2040
3000 CA Rotterdam, the Netherlands Email: e.heijnsdijk@erasmusmc.nl Phone:+31 10 7038842
Abstract
The European Randomized study of Screening for Prostate Cancer (ERSPC) has shown a 20% relative reduction in prostate cancer mortality after 16 years (rate ratio RR 0.80), but centers varied by attendance, screen interval, biopsy compliance, contamination in the control arm and treatments. We used a microsimulation model, calibrated to the ERSPC individual-level data, to predict influence of study features on the results. The relative reduction in prostate cancer mortality would have been somewhat larger with improved study features: increased attendance (90% attendance in all
volunteer-based and 70% in all population-based centers, resulting in RR 0.77), a 2-year screen interval (RR 0.75), and an 80% biopsy compliance (RR 0.79). The rate ratio would have been substantially lower with a 30% attendance (RR 0.92), 40% biopsy compliance (RR 0.90) or 100% contamination (RR 0.85). The variations in results by trial center may reflect differences in study design and performance and results of our simulations highlight the effect of quality indicators in PSA screening in different settings.
Patient summary: We evaluated the effect of various features of PSA screening on its effectiveness. The compliance to PSA testing and those having a biopsy after an elevated PSA substantially influence the prostate cancer mortality.
The European Randomized study of Screening for Prostate Cancer (ERSPC) has shown a 20% relative reduction in prostate cancer mortality after 16 years.[1] However, the results differed between centers, varying from a prostate cancer mortality rate ratio of 0.63 in Sweden to 0.99 in Italy. These differences could reflect variations in screening protocols, deviations from the quality criteria defined at the start of the trial, or variations in opportunistic screening and treatment.[2, 3]
In some centers the randomized population was volunteer-based (efficacy design) and attendance exceeded 90%. In the other centers a population-based effectiveness design was used with
randomization before consent and the attendance was 60%-70%. After the first round, the attendance was lower. The a priori defined quality criteria were 90% attendance for the volunteer-based centers and 70% for the population-volunteer-based centers. In France, the attendance was only around 30%.[4]
The screening interval in the trial was 4 years, except in Sweden (2 years). In Belgium, the interval between the first and second screen was 6 years.
The biopsy compliance varied between 40% (in third round in Italy) and 91% (Finland), whereas the quality criterion was that more than 80% of the men recommended for biopsy should be biopsied. The level of contamination in the ERSPC was estimated in some studies, ranging from 7%-40% per year across centers.[5-8] In Finland, 63% of the men in the control arm had a PSA test during the first 12 years of the trial.[9]
Treatment decisions were left to the treating physicians of each patient. This could have resulted in differences in treatment and therefore affect prostate cancer mortality reduction by trial center (Carlsson et al, submitted EUR Urol).
The aim of this study is to evaluate what the impact of the quality indicators has been on the overall ERSPC results.
For this analysis, we used the MISCAN prostate cancer model based on 16-year follow-up data of the ERSPC (excluding France). MISCAN is a stochastic model that simulates individual life histories. The model has been described extensively before and is described in the appendix.[10]
The output of the model was the prostate cancer incidence and mortality in both arms by year of follow-up. After calibration, the predicted prostate cancer mortality rate ratio at 16 years of follow-up was 0.79 versus 0.80 observed in the ERSPC (Appendix Figure 1).
The predicted prostate cancer mortality rate ratios for all scenarios are presented in Figure 1. The rate ratios varied from 0.92 for the 30% attendance scenario to 0.73 for the scenario, in which all men in the screen arm with loco-regional cancer were treated by radical prostatectomy. With inclusion of the French centers, the whole ERSPC study would show a prostate cancer mortality rate ratio of 0.88. The impact of screening on incidence and prostate cancer mortality by year of follow-up for all scenarios are presented in Appendix Figures 2-7.
The lower attendance after 4 years did not yet substantially influence prostate cancer mortality. An earlier study already found that the differences in prostate cancer mortality reduction between the centers could not be explained by the differences in attendance.[2] However, a very low attendance of 30%, as in France, would lead to a substantially lower incidence in the screening arm and also a lower prostate cancer mortality reduction.
The screening interval had a large impact on the incidence in the early years, but a smaller effect on prostate cancer mortality at 16 years. Possibly the influence on prostate cancer mortality will be larger with a longer follow-up.
The overall biopsy compliance was more than 80% in the first round of the ERSPC, decreasing slightly to 70%-75% in the third round, but the decline did not substantially influence the prostate cancer mortality. A very low biopsy compliance of 40% would have had a substantial impact on both incidence and mortality.
Contamination in the control arm substantially increased the incidence and reduced the prostate cancer mortality in the control arm, leading to a smaller prostate cancer mortality reduction by screening.
In a hypothetical situation, in which all men in the screen arm with a localized cancer would have been treated by radical prostatectomy, the prostate cancer mortality reduction would have increased to 27%. However, this would substantially diminish quality of life, due to complications and adverse effects of treatment.
This study has several limitations that involve the choice and uncertainty in the many parameters needed for the model. For example, SEER data on prostate cancer survival are used, instead of European data, because there is no large European dataset available for the survival in the period before screening started. We had to make assumptions on opportunistic screening. A sensitivity analysis could help in quantifying the uncertainty in the model parameters. However, for present analysis it is particularly complicated. When adjusting parameters, the point estimate of the prostate cancer mortality reduction in the base model will also change, complicating the interpretation of the results of the other scenarios. Also we modeled only a limited number of scenarios, based on the
ERSPC trial. Therefore, for example we didn’t asses the effect of different screen ages, since those were roughly similar in the ERSPC centers.
In conclusion, the ERSPC would have shown only a slightly larger prostate cancer mortality reduction if all centers had complied with the quality criteria for attendance or biopsy compliance, or if a 2-year screening interval had been used. In contrast, the prostate cancer mortality reduction could have been substantially lower if the attendance, biopsy compliance or contamination had been similar to the center with the lowest performance. The magnitude of the impact of the specific features estimated here may, together with information on overdiagnosis and overtreatment, be helpful in predicting the effectiveness of any (future) PSA testing program.
Acknowledgements
This publication was made possible by support of the EAU to the ERSPC study and by Grant Number U01 CA199338 from the National Cancer Institute as part of the Cancer Intervention and Surveillance Modeling Network (CISNET), which supported the underlying development of the simulation model utilized. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
Table 1. The inputs used in the base model and the 11 alternative scenarios
Base model Alternative scenario
Attendance By age, centre and round On average 82% in first round
1. 90% per round for NL, BEL, SPA, SWI and 70% per round for IT, FIN, SW 2. 30% per round
Interval By centre and round (2-6 year) 3. all 2 years 4. all 4 years 5. all 6 years Biopsy compliance By age, centre and round
On average 86% in first round
6. 80% 7. 40% Contamination mean of 0.3 tests per man in
control arm during the first 16 years of follow-up
8. mean of 0.6 tests per man 9. mean of 1.2 tests per man Curative treatment Treatment based on age and
stage
10. all men in screen arm and loco-regional stage receive radical prostatectomy
France not included 11. included (on average 30%
ERSPC including France all radical prostatectomy 1.2 tests contamination 0.6 tests contamination 40% biopsy compliance 80% biopsy compliance screen interval 6 years screen interval 4 years screen interval 2 years 30% attendance 90% and 70% attendance base model
0.7 0.8 0.9 1.0
Figure 1. Predicted prostate cancer mortality rate ratio in the ERSPC trial after 16 years of follow-up for the various scenarios.
References
[1] Hugosson J, Roobol MJ, Mansson M, Tammela TLJ, Zappa M, Nelen V, et al. A 16-year Follow-up of the ERSPC trial. Eur Urol 2019, Feb 26.
[2] Hakama M, Moss SM, Stenman UH, Roobol MJ, Zappa M, Carlsson S, et al. Design-corrected variation by centre in mortality reduction in the ERSPC randomised prostate cancer screening trial. J Med Screen. 2017;24:98-103.
[3] Roobol MJ, Schröder FH. European Randomized Study of Screening for Prostate Cancer: achievements and presentation. BJU Int. 2003;92 Suppl 2:117-22.
[4] Schröder FH, Hugosson J, Roobol MJ, Tammela TL, Zappa M, Nelen V, et al. Screening and prostate cancer mortality: results of the European Randomised Study of Screening for Prostate Cancer (ERSPC) at 13 years of follow-up. Lancet. 2014;384:2027-35.
[5] Bokhorst LP, Bangma CH, van Leenders GJ, Lous JJ, Moss SM, Schroder FH, et al. Prostate-specific antigen-based prostate cancer screening: reduction of prostate cancer mortality after correction for nonattendance and contamination in the Rotterdam section of the European Randomized Study of Screening for Prostate Cancer. Eur Urol. 2014;65:329-36.
[6] Ciatto S, Zappa M, Villers A, Paez A, Otto SJ, Auvinen A. Contamination by opportunistic screening in the European Randomized Study of Prostate Cancer Screening. BJU Int. 2003;92 Suppl 2:97-100. [7] Lujan M, Paez A, Pascual C, Angulo J, Miravalles E, Berenguer A. Extent of prostate-specific antigen contamination in the Spanish section of the European Randomized Study of Screening for Prostate Cancer (ERSPC). Eur Urol. 2006;50:1234-40.
[8] Otto SJ, van der Cruijsen IW, Liem MK, Korfage IJ, Lous JJ, Schroder FH, et al. Effective PSA
contamination in the Rotterdam section of the European Randomized Study of Screening for Prostate Cancer. Int J Cancer. 2003;105:394-9.
[9] Kilpeläinen TP, Pogodin-Hannolainen D, Kemppainen K, Talala K, Raitanen J, Taari K, et al. Estimate of Opportunistic Prostate Specific Antigen Testing in the Finnish Randomized Study of Screening for Prostate Cancer. J Urol. 2017;198:50-7.
[10] de Koning HJ, Gulati R, Moss SM, Hugosson J, Pinsky PF, Berg CD, et al. The efficacy of prostate-specific antigen screening: Impact of key components in the ERSPC and PLCO trials. Cancer.
2018;124:1197-206.
Appendix
The impact of quality in PSA testing: differences between ERSPC centres
Eveline A.M. Heijnsdijk, Jan Adolfsson, Anssi Auvinen, Monique J. Roobol, Jonas Hugosson, Harry J. de Koning.
The MISCAN model
MISCAN is a stochastic model that simulates individual life histories. The model has been described extensively before. [1, 2] In short, in the MISCAN model, disease progresses through a sequence of states defined by stage (clinical T stage 1, 2 and 3) and grade (Gleason 6 or smaller, Gleason 7 and Gleason 8 or larger). In each state, there is a probability of clinical detection and, dependent on the screen sensitivity and attendance, a probability of screen detection. The model has been calibrated to
ERSPC data by adjusting disease progression rates and the PSA test sensitivity against both the incidence and stage distributions of clinically detected cancers in the control arm, as well as the screen-detected and interval cancers in the screen arm. We used enrollment patterns, screen
attendance, and biopsy compliance by age to model the number of screens and biopsies in the screen arm. PSA testing in the control arm (contamination) was simulated using a model described
elsewhere. [3]
Survival in the absence of screening was modeled using a Poisson regression model fit to
Surveillance, Epidemiology, and End Results (SEER) data for untreated cases diagnosed in 1983-1986, just prior to the advent of PSA screening. [4] This baseline survival was improved for localized cases who received radical prostatectomy, or radiation therapy in combination with hormone therapy, using a hazard ratio of 0.62 and localized cases who received radiation therapy using a hazard ratio of 0.7. [5] Distributions of treatments were based on age, Gleason score, and stage, as observed in the ERSPC trial. Other-cause survival was generated using country specific life tables for men in the different centres.
Modeling screening benefit
The mortality benefit of PSA screening was modeled as a cure probability that depended on the lead time (years by which detection of the cancer is advanced by screening compared to the clinical situation) and was implemented only for screen-detected, non-metastatic, and non-overdiagnosed cases as cure probability = 1 – exp (-cure parameter x lead time). Thus, the probability of cure increases with lead time. Cured men were assigned to die at their independently generated date of other-cause death. Men who were not cured died at the same time they would have died if they had not been screened. In a previous study modeling the PLCO trial, models developed by two groups substantially over-projected observed prostate cancer mortality despite closely reproducing incidence and stage and grade patterns.[6]. Therefore, we included a baseline survival hazard ratio to improve the baseline survival, to reflect improvements in disease management since the period 1983-1986 beyond screening or primary treatment. [3]
In the previous study, we jointly calibrated the cure parameter and the baseline survival hazard ratio were calibrated to both the ERSPC and PLCO.[3] For this analysis, we updated these two parameters using the ERSPC 16 year follow-up data. The parameters used were a cure parameter of 0.16 and a baseline survival hazard ratio of 0.74.
Appendix Figure 1. The observed cumulative prostate cancer incidence and prostate cancer mortality (triangles) compared with the model predictions (lines) for the screen arm (black) and the control arm (grey). 0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 140 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Year of follw-up C u m u la ti ve p ro st at e c an ce r d e at h s p e r 1 0 0 0 m e n
Appendix Figure 2. Cumulative prostate cancer incidence and mortality by year of follow-up for highest and lowest centre-specific attendance rates
0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Year of follow-up C u m u la ti ve m o rt al it y p e r 1 0 0 0 m e n
Screen arm, base model Control arm, base model Screen arm, 30% attendance
Appendix Figure 3. Cumulative prostate cancer incidence and mortality by year of follow-up by screen interval 0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Year of follow-up C u m u la ti ve m o rt al it y p e r 1 0 0 0 m e n
Screen arm, base model Control arm, base model Screen arm, screen interval 6 years
Screen arm, screen interval 4 years Screen arm, screen interval 2 years
Appendix Figure 4. Cumulative prostate cancer incidence and mortality by year of follow-up for highest and lowest centre-specific biopsy compliance
0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Year of follow-up C u m u la ti ve m o rt al it y p e r 1 0 0 0 m e n
Screen arm, base model Control arm, base model Screen arm, 40% biopsy compliance
Appendix Figure 5. Cumulative prostate cancer incidence and mortality by year of follow-up by contamination in the control arm
0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 Year of follow-up C u m u la ti ve m o rt al it y p e r 1 0 0 0 m e n
Screen arm, base model Control arm, base model
Control arm, 0.6 tests contamination
Control arm, 1.2 tests contamination
Appendix Figure 6. Cumulative prostate cancer incidence and mortality by year of follow-up when all men in screen arm receive radical prostatectomy
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Year of follow-up C u m u la ti ve m o rt al it y p e r 1 0 0 0 m e n
Screen arm, base model Control arm, base model Screen arm, all men radical prostatectomy
Appendix Figure 7. Including France in the ERSPC results 0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 Year of follow-up C u m u la ti ve in ci d e n ce p e r 1 0 0 0 m e n 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 Year of follow-up C u m u la ti ve d e at h s p e r 1 0 0 0 m e n
Screen arm, base ERSPC model Control arm, base ERSPC model Screen arm, including France Control arm, including France
References
[1] Heijnsdijk EA, de Carvalho TM, Auvinen A, Zappa M, Nelen V, Kwiatkowski M, et al. Cost-effectiveness of prostate cancer screening: a simulation study based on ERSPC data. J Natl Cancer Inst. 2015;107:366.
[2] Heijnsdijk EA, Wever EM, Auvinen A, Hugosson J, Ciatto S, Nelen V, et al. Quality-of-life effects of prostate-specific antigen screening. N Engl J Med. 2012;367:595-605.
[3] de Koning HJ, Gulati R, Moss SM, Hugosson J, Pinsky PF, Berg CD, et al. The efficacy of prostate-specific antigen screening: Impact of key components in the ERSPC and PLCO trials. Cancer. 2018;124:1197-206.
[4] Albertsen PC, Hanley JA, Fine J. 20-year outcomes following conservative management of clinically localized prostate cancer. JAMA. 2005;293:2095-101.
[5] Etzioni R, Gulati R, Tsodikov A, Wever EM, Penson DF, Heijnsdijk EA, et al. The prostate cancer conundrum revisited : Treatment changes and prostate cancer mortality declines. Cancer. 2012;118:5955-63.
[6] Gulati R, Tsodikov A, Wever EM, Mariotto AB, Heijnsdijk EA, Katcher J, et al. The impact of PLCO control arm contamination on perceived PSA screening efficacy. Cancer Causes Control. 2012;23:827-35.
