• No results found

Early diagnosis and treatment of prostate cancer

N/A
N/A
Protected

Academic year: 2021

Share "Early diagnosis and treatment of prostate cancer"

Copied!
58
0
0

Loading.... (view fulltext now)

Full text

(1)

UMEÅ UNIVERSITY MEDICAL DISSERTATIONS

New series No.1397 | ISSN No. 0346-6612 | ISBN No. 978-91-7459-134-7

___________________________________________

Early diagnosis and treatment

of prostate cancer

Observational studies in

the National Prostate Cancer Register of Sweden

and the Västerbotten Intervention

Project

Benny Holmström

Department of Surgical and Perioperative Sciences, Urology and Andrology, Umeå University, Umeå Centre for Research and Development,

Uppsala University/County Council of Gävleborg, Gävle

(2)

Department of Surgical and Perioperative Sciences Urology and Andrology

Umeå University 901 85 Umeå Sweden

Centre for Research and Development

Uppsala University/County Council of Gävleborg 801 87 Gävle

Sweden

Responsible publisher under Swedish law - the Dean of the Faculty of Medicine This work is protected by the Swedish Copyright Legislation (Act 1960:729) Copyright © Benny Holmström, benny.k.holmstrom@gmail.com

ISSN: 0346-6612

ISBN: 978-91-7459-134-7

(3)
(4)

ABSTRACT

Background: Prostate-specific antigen (PSA) testing has caused a steep increase in the incidence of prostate cancer, especially the incidence of localised low risk disease. In order to decrease the overdiagnosis accompanied by PSA testing, analysis of inherited genetic variants have been suggested as potential tools for clinical assessment of disease risk. With the aim of minimizing overtreatment and postpone side-effects of curative treatment for low risk prostate cancer, active surveillance, a treatment strategy with initial surveillance and deferred radical prostatectomy at the time of progression has evolved.

Aim: The aim of this thesis was to study the validity of PSA (paper I) and inherited genetic variants (paper II) for early diagnosis of prostate cancer, to assess the extent of PSA testing in Sweden (paper III), and to study the safety of deferred radical prostatectomy in localised low to intermediate risk prostate cancer (paper IV).

Methods: The study designs were i) case-control studies nested within the Västerbotten intervention project (paper I and II), ii) observational study in the Swedish Cancer Register (paper III), and iii) observational study in the NPCR Follow-up study (paper IV).

Results: PSA had a high validity in predicting a prostate cancer diagnosis with an area under the receiver operating characteristics (ROC) curve of 0.86 (95% CI, 0.84 to 0.88). A combined test, including PSA, the ratio of free to total PSA, and 33 single nucleotide polymorphisms (SNPs) in a genetic risk score, increased the area under curve to 0.87 (95% CI, 0.85 to 0.89). The estimated uptake of PSA testing among men aged 55 to 69 years increased from zero to 56% between 1997 and 2007 and there were large variations in the uptake of PSA testing between counties in Sweden. After a median follow-up time of eight years there was no significant difference in presence of any one or more adverse pathology features or prostate cancer specific mortality after primary compared to deferred radical prostatectomy in localised low to intermediate risk prostate cancer.

Conclusions: Results from these studies indicate that PSA and the hitherto identified SNPs are not suitable biomarkers in single-test prostate cancer screening. It is possible to estimate the uptake of PSA testing on a population level. Initial surveillance and deferred radical prostatectomy represent a feasible treatment strategy in localised low to intermediate risk prostate cancer.

(5)

TABLE OF CONTENTS

ABSTRACT ... 4 ABBREVIATIONS ... 7 LIST OF PAPERS ... 8 THESIS AT A GLANCE ... 9 1 INTRODUCTION ... 10 1.1 Prostate cancer ... 10

1.2 Prostate cancer incidence ... 10

1.3 Prostate-specific antigen – PSA ... 12

1.4 Single nucleotide polymorphism – SNP ... 12

1.5 Treatment of localised low risk prostate cancer ... 13

2 AIMS ... 15

3 MATERIALS AND METHODS ... 16

3.1 Västerbotten Intervention Project (VIP) ... 16

3.2 The Swedish Cancer Register ... 16

3.3 The Swedish Cause of Death Register ... 17

3.4 National Prostate Cancer Register (NPCR) of Sweden ... 17

3.5 Biochemical analysis ... 17

3.6 Genotyping methods ... 18

3.7 Uptake of PSA testing ... 18

3.8 Statistical analysis ... 19

3.8.1 Paper I and II ... 19

3.8.2 Paper III ... 20

3.8.3 Paper IV ... 20

3.9 Ethical considerations ... 20

4 RESULTS AND DISCUSSION ... 21

4.1 PSA and SNP – Paper I and II ... 21

(6)

4.1.2 Discussion ... 27

4.2 Uptake of PSA testing – Paper III ... 29

4.2.1 Results ... 29

4.2.2 Discussion ... 29

4.3 Primary versus deferred radical prostatectomy – Paper IV ... 32

4.3.1 Results ... 32

4.3.2 Discussion ... 35

4.4 Strengths and limitations ... 37

4.4.1 Evaluation of validity of PSA and SNPs (Papers I and II) ... 37

4.4.2 Estimation of uptake of PSA testing (Paper III) ... 38

4.4.3 Primary versus deferred radical prostatectomy (Paper IV)... 38

5 FUTURE PERSPECTIVES ... 40

6 SUMMARY ... 42

7 SAMMANFATTNING PÅ SVENSKA ... 44

8 ACKNOWLEDGEMENTS ... 46

(7)

7

ABBREVIATIONS

CI Confidence interval CV Coefficient of variation DNA Deoxyribonucleic acid

NPCR National Prostate Cancer Register OR Odds ratio

PCR Polymerase chain reaction PSA Prostate-specific antigen

ROC Receiver operating characteristics SNP Single nucleotide polymorphism TNM Tumour Node Metastasis

(8)

8

LIST OF PAPERS

This thesis is based on the results from the following papers, which are referred to in the text by their Roman numerals:

I. Holmström B, Johansson M, Bergh A, Stenman UH, Hallmans G, Stattin P. Prostate specific antigen for early detection of prostate cancer: longitudinal study. BMJ 2009;339:b3537

II. Johansson M, Holmström B, Hinchliffe SR, Bergh A, Stenman UH, Hallmans G, Wiklund F, Stattin P. Combining 33 genetic variants with prostate-specific antigen for prediction of prostate cancer: Longitudinal study. Int J Cancer 2011 Feb 15. [Epub ahead of print] (doi: 10.1002/ijc.25986)

III. Jonsson H, Holmström B, Duffy SW, Stattin P. Uptake of prostate-specific antigen testing for early prostate cancer detection in Sweden. Int J Cancer 2010 Dec 10. [Epub ahead of print]

IV. Holmström B, Holmberg E, Egevad L, Adolfsson J, Johansson JE, Hugosson J, Stattin P. Outcome of primary versus deferred radical prostatectomy in the National Prostate Cancer Register of Sweden Follow-up study. J Urol 2010;184:1322-27.

(9)

9

THESIS AT A GLANCE

Paper Design Endpoints Main findings

I VIP, nested

case-control Incident PCa (n=540), low risk (n=355) and high risk (n=185) tumours

No cut-off of PSA met the standards of a screening test in terms of likelihood ratios.

II VIP, nested

case-control Incident PCa (n=520), low risk (n=341) and high risk (n=179) tumours

A marginal improvement in prediction of PCa was achieved by adding information on 33 common genetic variants.

III Nation-wide,

observational Uptake of PSA testing The estimated cumulated uptake of PSA testing increased from zero to 56% between 1997 and 2007. IV NPCR,

observational Primary RP (n=2344) and deferred RP (n=222)

There was no significant difference in presence of any one or more of adverse pathology features or in PCa specific mortality after primary compared to deferred RP.

VIP, Västerbotten Intervention Project; PCa, prostate cancer; PSA, prostate-specific antigen; NPCR, National Prostate Cancer Register; RP, radical prostatectomy

(10)

10

1 INTRODUCTION

1.1 Prostate cancer

Prostate cancer is, except for skin cancer, the most common cancer among men in developed countries and accounts for almost one third of cancers diagnosed in US males1. In Sweden prostate cancer is currently diagnosed in between 9,000 and 10,000 men each year and accounts for approximately 2,500 deaths per year2. The prevalence of prostate cancer increase by age, and depending on how meticulous the examinations for detecting prevalent cases are done, different numbers will be achieved3,4. In an autopsy study from the US, prevalent prostate cancer was found among 8% in men in their 20ies to 30ies to over 80% in men in their 70ies (Table 1)5. Not all prevalent cases of prostate cancer are clinically significant, i.e. pose a threat to life, since considerably fewer men die from prostate cancer each year compared to the high prevalence4. Analysis of the distribution of sites of origin of prostate cancer has shown that approximately 70% of cancers develop in the peripheral zone, 20% in the transition zone, and 10% in the central zone of the prostate gland6.

Table 1. Prevalence of prostate cancer reported from autopsies5.

Age US White (%) US Black (%)

21-30 8 8 31-40 31 31 41-50 37 43 51-60 44 46 61-70 65 70 71-80 83 81

1.2 Prostate cancer incidence

During the past two decades the prostate cancer incidence has increased dramatically in the Western world, including the Nordic countries7. In the US the incidence peaked in 1992 with an age standardised prostate cancer incidence of about 240 per 100 0001. In Sweden the age standardised prostate cancer incidence increased to 236 per 100 000 in 2004, and the greatest increase was seen for low risk prostate cancer (Figure 1)2,8. The total number of incident prostate cancer cases increased from 5,946 in 1997 to 10,317 in 2009 (Figure 2)2. During the same time the age standardised incidence of patients diagnosed with metastatic disease in Sweden decreased9. However, there are considerable differences in the prostate cancer incidence between counties in Sweden. It has been suggested that the increased incidence is due to an increased uptake of PSA testing, both as a part of the work-up of lower urinary tract symtoms but also due to

(11)

11

unorganized PSA screening activities7,9. However, other reasons such as better diagnostic tools (ultrasound guided prostate biopsies, more biopsies per biopsy-session etc.) might be a part of the explanation.

Figure 1. The age-standardized incidences between 1991 and 2008 of men diagnosed with incident low risk prostate cancer in Sweden8. Since the NPCR did not cover all regions in

Sweden until 1998, estimations were made for the years between 1991 to 1997 by utilizing available data and the background population. Low risk prostate cancer defined according to the National Comprehensive Cancer Network. Practice Guidelines in Oncology-Version.1.2010. Prostate cancer.

Figure 2. The annual numbers of incident prostate cancer cases diagnosed in Sweden between 1970 and 20092. 0 50 100 1991 1993 1995 1997 1999 2001 2003 2005 2007 A ge-st andar di zed in ci denc e per 100 000 0 2000 4000 6000 8000 10000 12000 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009

(12)

12

1.3 Prostate-specific antigen – PSA

PSA is one of several tissue kallikreins, and PSA is produced by the epithelium in the prostate glandular acini10. The main function of PSA is to liquefy the seminal clot in order to increase the motility of the spermatozoa11. The basement membrane in the wall of the acini leaks a small fraction of the kallikreins enabling measurement of tissue kallikreins in circulating blood.

PSA was described in 1979 by Wang and coworkers12, and in 1987 Stamey et al showed that PSA performed better as a serum marker for prostate cancer than prostatic acid phosphatase13. The Food and Drug Administration (FDA) approved PSA for use in monitoring of prostate cancer in 1986, and in 1994 FDA approved the use of PSA for early diagnosis of prostate cancer14. Since the mid 1990ies, PSA has been widely used for work-up of lower urinary tract symtoms, monitoring of prostate cancer as well as for early diagnosis of localised prostate cancer15.

PSA is the most commonly used serum marker in oncology11. Evaluation of the validity of PSA as a diagnostic tool has mostly been performed in cross-sectional studies with prostate biopsies performed in men with PSA values over a certain cut-off16-19. Studies with such design are influenced by verification bias and cannot make accurate estimates of sensitivity since men with PSA values below the cut-off does not undergo biopsies20. However, it is possible to estimate sensitivity in cross-sectional studies in which all men undergo prostate biopsies regardless of PSA level21. It is also possible to accurately estimate sensitivity and specificity in longitudinal studies were blood samples are collected and stored many years prior to diagnosis and were the prediagnostic PSA value are unknown and does not trigger any diagnostic activity22-27.

Previous evaluations of PSA have focused on the validity of PSA in predicting a prostate cancer diagnosis. In view of the recently published randomised controlled prostate cancer screening trials, it is also essential to evaluate if PSA meet the standard of a screening test. Likelihood ratios have been proposed as powerful tools for evaluating the clinical usefulness of a screening test since they are not affected by disease prevalence28.

1.4 Single nucleotide polymorphism – SNP

Since the start of the human genome project 20 years ago, the human genome, consisting of 3 billion base pairs, have been explored29-31. New technological advancements since the start of the project have made it feasible to explore the human genome in a fast and economically feasible

(13)

13

way. In such genome-wide association studies alterations in the nucleotides, i.e. adenine, cytosine, guanine, and thymine encoding the DNA, have been identified and independently replicated for a number of disorders32. The early success triggered hope for identifying subjects with elevated risk of disease, and several studies have subsequently used genetic risk scores in prediction of prostate cancer33-37. However, common genetic variants typically confer small relative risks individually, and the discriminative performance reported in previous studies has been limited, even when coupled with information on family history of prostate cancer.

1.5 Treatment of localised low risk prostate cancer

The management of localised prostate cancer is still controversial due to lack of randomised controlled trials comparing the different treatment options available38. Early studies indicated a lower rate of progression after surgery than after external radiotherapy39, but no gain in overall survival after more than 20 years of follow-up, as compared with primary expectant management40,41. Expectant management or watchful waiting means that the patient is left untreated until the time of progression or development of symtoms of disease and that the treatment instigated at that point is non-curative42.

Radical prostatectomy as a surgical procedure was first described by Billroth in the mid 19th century43. However, the high frequency of morbidity, i.e. mostly postoperative erectile dysfunction and urinary incontinence, but also peri- and postoperative mortality44, delayed the introduction of surgical treatment for localised prostate cancer. After the introduction in the early 1980ies of a nerve-sparing technique45,46, decreasing the side-effects of treatment, the utilization of surgery as a treatment option became more widespread. However, at that time there was no evidence in support of a benefit from radical prostatectomy compared to watchful waiting in terms of survival. In 2002 a randomised controlled trial was reported, in which 695 men with clinically localised prostate cancer had been randomly assigned to radical prostatectomy (n = 347) or watchful waiting (n = 348). This landmark study showed a modest but statistically significant survival benefit for men who underwent radical prostatectomy47. Since then, radical prostatectomy has become widely used for treatment of localised prostate cancer48.

Due to the early findings of Chodak et al and Albertsen et al that men with well-differentiated prostate cancer have a prostate cancer specific survival rate of 80 to 90% after 20 years of follow-up49,50, a treatment strategy with

(14)

14

deferred definite curative treatment, i.e. active surveillance, has evolved for the increasing amount of low risk prostate cancers diagnosed51. A recently published population-based, nation-wide study showed that men with localised low risk prostate cancer put on surveillance have a low risk of dying from prostate cancer on a 10-year basis52. However, the long-time safety in deferred radical prostatectomy for low risk prostate cancer is yet uncertain.

(15)

15

2 AIMS

The aim of this thesis was to study early detection and surgical treatment of localised low to intermediate risk prostate cancer.

Specific aims were to:

 Evaluate if the validity of PSA in predicting a future prostate cancer diagnosis meet the formal standard of a screening test (paper I).

 Evaluate if analysis of PSA, the ratio of free to total PSA, and a genetic risk score in a combined test increase the validity in order to meet the standard of a screening test (paper II).

 Evaluate if it is possible to estimate the uptake of PSA testing on a population level (paper III).

 Evaluate if deferred radical prostatectomy represents a safe treatment strategy in men diagnosed with localised low to intermediate risk prostate cancer (paper IV).

(16)

16

3 MATERIALS AND METHODS

3.1 Västerbotten Intervention Project (VIP)

In 1985 an intervention programme started in the Norsjö municipality, Västerbotten, Sweden, aiming at decreasing the cardiovascular morbidity and mortality among the inhabitants53. From 1991 and onwards all residents in the county of Västerbotten (population size approximately 258,000 in 2007) are invited to participate in a health survey in the calendar year in which they become 40, 50 or 60 years old. At initial recruitment, subjects are asked to complete a self-administered comprehensive questionnaire including demographic, medical and lifestyle information, and subjects are also asked to donate a blood sample which is cryopreserved in -80°C at the Umeå University Medical Biobank54. From inception through 2011-02-11, 87,520 unique individuals have participated and donated blood samples to the Medical Biobank (Göran Hallmans, personal communication). By January 2006 a total of 654 incident prostate cancer cases were identified by linkage of the VIP cohort to the regional cancer register. Out of those, 540 (83%) cases had prospectively collected plasma samples available for PSA analysis in paper I, of which 520 (96%) cases had genomic DNA available for genotyping in paper II. High risk prostate cancer was defined as local clinical tumour stage T3 or T4, Gleason score 8 or higher or WHO grade 3, presence of lymph node metastasis, bone metastasis or serum PSA levels above 20 ng/mL. No formal screening program for prostate cancer has been, or is currently, in operation in the source population. For each case, two controls that were alive and free of cancer at the time of diagnosis of the index case, were randomly selected within sets matched to the index case for age (+/- 6 months) and date of recruitment (+/- 2 months). There were 1,034 and 988 individually matched controls in paper I and paper II, respectively.

3.2 The Swedish Cancer Register

Since 1958 all incident cases of cancer have been recorded in the nation-wide Swedish Cancer Register which contains basic data, including the per individual unique personal identity number, date of diagnosis, municipality and county of residence of the patient at diagnosis, and the type and site of each new tumour9. According to regulations issued by the National Board of Health and Welfare, all physicians in hospitals or other establishments for medical treatment under public or private administration in Sweden must report all new cases of cancer. Pathologists and cytologists separately report cases with a cancer diagnosis. Thus, for the majority of cases the Swedish Cancer Register is notified twice, i.e. by the clinical and pathology departments. After quality control and coding, the data are transferred to the

(17)

17

national cancer register, and the overall capture rate has been estimated to 96.3 % for all tumours and higher among solid tumours in subjects 70 years or younger55.

3.3 The Swedish Cause of Death Register

The first efforts to introduce a nation-wide register for the Swedish population statistics was made in 1749 when the clergymen was imposed the duty of registering the causes of death. The Swedish Cause of Death Register in its present configuration has been in operation since 1961 and contains data on personal identity number, date of death, underlying cause of death, municipality and county of residence of the person at time of death2. Registration of date of death and underlying cause of death is mandatory and the validity of the Swedish Cause of Death Register has been demonstrated to be high, 88% of deaths attributed to prostate cancer among men diagnosed with localised disease were also attributed to prostate cancer in a re-examination of medical records56.

3.4 National Prostate Cancer Register (NPCR) of Sweden Approximately 97% of all incident prostate cancer cases in the Swedish Cancer Register are also registered in the National Prostate Cancer Register (NPCR) of Sweden. The register contains information on diagnosing unit, date of diagnosis, cause of diagnosis, tumour grade, tumour stage according to the TNM classification in force, serum PSA levels at diagnosis and primary treatment either performed or planned up to six months after diagnosis9. Until 2007, treatment coded “expectancy” included both active surveillance and watchful waiting. Since 2007 also information on prostate size, amount and total length of core biopsies and length of cancer in the cores are registered as well as the extent of nerve sparing on radical prostatectomy. The NPCR started in 1996 and included all regions in Sweden from 1998 and onwards.

3.5 Biochemical analysis

All baseline assays of PSA and free PSA included in paper I and II were analysed at the Department of Clinical Chemistry, Helsinki University Central Hospital, Helsinki, Finland, by using Wallac Delfia assays (AutoDelfia, Wallac Oy, Turku, Finland). The intra-assay and inter-assay CVs were in the range 2-4% at PSA levels between 0.2 and 100 ng/mL. The laboratory personnel were blinded to case-control status and cases and their matched controls were analysed in the same batch.

(18)

18

The assays of serum PSA levels in blood drawn shortly before the diagnostic biopsies utilized either Hybritech Tandem-R (Hybritech Inc, San Diego, CA) or the IMx PSA assay (Abbot Laboratories, Abbot Park, IL), and values had been recorded in the NPCR. The coefficient of correlation between the two assays was 0.990 (IMx value = (1.22  Tandem value)2.80)57.

3.6 Genotyping methods

In paper II the genotyping was performed at the Mutation Analysis Facility in Huddinge, Stockholm, Sweden, using the matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (Sequenom Inc., San Diego, CA, USA)58. The order of DNA samples from cases and controls was randomised on PCR plates in order to assure that an equal proportion of cases and controls could be analysed simultaneously. The laboratory personnel were kept blinded to case-control status and quality control was achieved by analysing the concordance rates between duplicate samples. SNPs were analysed for deviations from Hardy-Weinberg equilibrium using χ2 statistics. The rs445114 SNP had 212 missing genotype calls and did not conform to Hardy-Weinberg equilibrium (p=1x10-5), but the SNP was still found to be associated with prostate cancer risk (p for trend=0.005) in accordance with the original genome-wide association studies59-63, and was therefore retained in the genetic risk score. Excluding rs445114 genotyping call rates ranged between 96.2% and 100% (average call rate: 99.7%). Duplicate sample concordance rate was 100%.

3.7 Uptake of PSA testing

There are no national or regional databases recording the PSA testing in Sweden. In order to estimate the uptake of PSA testing from the incidence pattern among Swedish counties, it is possible to utilize data from randomised population-based PSA screening trials. In such trials men are randomised to a screening group or to a control group. In the screening group the number of screening visits are known as well as the resulting prostate cancer incidence, i.e. incidence of prostate cancer in a screened population. Men in the control group are not subjected to screening and the prostate cancer incidence is also known for those men, i.e. incidence of prostate cancer in an unscreened population.

Data from a published randomised population-based PSA screening trial, the Göteborg part of the ERSPC trial64, were extracted in order to estimate the uptake of PSA testing in Sweden. In this study, which started in 1995, a total of 20,000 men aged 50 to 64 years were randomised in a 1:1 ratio to a

(19)

19

screening group offered biennial screening or to a control group64. We utilized results from four screening rounds, i.e. eight years, in our analysis.

Since there were hardly any PSA testing during the 1980ies, we used historical data on the prostate cancer incidence between 1980 and 1990 for prediction of the prostate cancer incidence without any PSA testing.

3.8 Statistical analysis

Throughout the studies included in this thesis standard statistical methods were used, such as linear and logistic regression, in order to investigate various relationships. P-values lower than 0.05 were considered statistically significant. OR was considered statistically significant if the corresponding p-value was below 0.05.

3.8.1 Paper I and II

In paper I, conditional logistic regression was used to calculate OR of prostate cancer for various PSA cut-offs. Because of the individually matched case-control design, demographic factors were automatically accounted for. Specificity and sensitivity estimates were calculated for a series of total PSA cut-offs along with positive and negative likelihood ratios. The positive likelihood ratio was calculated according to [sensitivity/(1-specificity)], and the negative likelihood ratio was calculated as [(1-sensitivity)/specificity]. In papers I and II, ROC analysis was used to generate area under curve estimates (along with Wald 95% CI) as measures of the overall validity in prediction of prostate cancer diagnosis for the full study group, and for subgroups stratified according to time from blood draw to diagnosis, age at blood draw, and low/high risk cancer.

In paper II, unconditional logistic regression was used to generate ORs and probability estimates for ROC analysis, and here demographic factors were accounted for by additional adjustments for age at recruitment in three groups and year of recruitment. The original genome-wide association studies were used to define the risk alleles59-63, and the genetic risk score was created according to the number of risk alleles carried. Six risk models were evaluated: no risk variables (including demographic variables only); total PSA; the ratio between free and total PSA; the logarithm of total PSA and the ratio between free and total PSA; the genetic risk score; and finally the full model including the genetic risk score, and the logarithm of total PSA and the ratio between free and total PSA.

(20)

20

3.8.2 Paper III

Estimates of the prostate cancer incidence trends without any PSA testing were made with a log-linear Poisson model with calendar year (continuous) and county as covariates. The prostate cancer incidence between 1980 and 1990, when there were almost no PSA testing, were utilized for the estimations. Separate intercepts were used for all counties but with a common slope for calendar year.

3.8.3 Paper IV

In paper IV, unconditional logistic regression was used to calculate OR for different categories of adverse pathology, adjusting for age, PSA, Gleason score, T-classification and year of diagnosis. Observation time was calculated from date of diagnosis and time at risk was calculated from date of radical prostatectomy. Difference in distribution of the variables among men who underwent primary or deferred radical prostatectomy were tested by the use of χ2 statistics.

3.9 Ethical considerations

Written informed consent had been obtained from all study participants at baseline in paper I, II and IV. Since the study for paper III did not contain any diagnostic activity, sampling of specimens or anything else that affected the study participants, informed consent was not needed. The studies in paper I-III were approved by the research ethics committee at Umeå University Hospital, Umeå, Sweden, and the study in paper IV was approved by the research ethics committee at Göteborg University, Göteborg, Sweden.

(21)

21

4 RESULTS AND DISCUSSION

4.1 PSA and SNP – Paper I and II 4.1.1 Results

Characteristics of study subjects and prostate cancer cases in paper II are presented in Table 2. The loss of eligible cases (n=20) and controls (n=46) (because of missing DNA for those samples) for analysis in the cohort in paper II did not materially change the characteristics of the study subjects. The median PSA levels at baseline was 1.1 ng/mL among controls and 3.6 ng/mL among cases. The distribution of PSA levels and risk alleles among cases and controls in paper I and II is shown in Figure 3 and Figure 4, respectively. In paper II, 17 SNPs displayed associations (Ptrend<0.10) with prostate cancer risk in accordance with the original genome-wide association studies59-63, and two SNPs displayed OR estimates opposite to those in the original genome-wide association studies (Ptrend: 0.22-0.50).

Figure 3. Distribution of plasma PSA in cases and controls. The histogram shows the observed distribution of the logarithm of PSA in cases and controls. Solid lines indicate the frequency functions of the normal distribution of the logarithm of PSA according to the average and standard deviations in cases and controls. Figure reproduced from Holmström et al., paper I66

(22)

22

Table 2. Characteristics of study subjects and prostate cancer cases in paper II. Table reproduced from Johansson et al., paper II65.

Continuous variables, Median (5th-95th percentiles)

Cases (n=520) Controls (n=988)

Age at blood draw 59 (49 - 60) 59 (49 - 60)

PSA at blood draw (ng/mL) 3.6 (1.2 - 21) 1.1 (0.35 - 5.2)

Case characteristics Cases

Age at diagnosis (years) 64.0 (54.0 – 71.0)

Time between blood draw and diagnosis (years) 7.0 (1.0 – 13.0)

PSA at diagnosis (ng/mL) 11.0 (4.0 - 145) Discrete variables—No (%) Stage: T1a,b 23 (5) T1c 232 (47) T2 174 (35) T3 60 (12) T4 6 (1) TX 25

Lymph node metastasis:

N0 160 (92) N1 14 (8) NX 346 Bone metastasis: M0 308 (89) M1 40 (11) MX 172 Gleason score: 2-6 277 (63) 7 126 (29) 8-10 39 (9) Missing 78

WHO grade where Gleason score missing:

I 25 (39) II 25 (39) III 14 (22) Missing 14 PSA at diagnosis (ng/mL) <10 216 (44) 10-20 143 (29) 20-50 73 (15) >50 62 (13) Missing 26

High/low risk prostate cancer1

Low risk 341 (66)

High risk 179 (34)

1) High risk defined as clinical local tumour stage T3 or T4, lymph node metastasis (N1), bone metastasis (M1), Gleason score ≥ 8, WHO grade III, or serum levels of PSA at diagnosis > 20 ng/mL; low risk defined as absence of all of these factors.

(23)

23

Figure 4. Distribution of risk alleles for cases and controls with projected ORs. Black diamonds represent ORs relative having 28 risk allele (reference category) and were calculated based on the overall per risk increasing allele trend estimate and gray shading indicates corresponding 95% confidence interval. Figure reproduced from Johansson et al., paper II65.

Table 3. OR with 95% CI of prostate cancer during follow-up for different levels of PSA. Reproduced from Holmström et al., paper I66.

PSA (ng/mL) OR (95% CI)* <1 1.0 (ref) 1-2 9.1 (5.0–16.5) 2-3 23.3 (12.3-43.9) 3-4 43.9 (22.1-87.3) 4-10 68.1 (35.2-130.6) >10 239.5 (89.3-642.3)

*) Calculated with the use of conditional logistic regression

In paper I, the OR for a prostate cancer diagnosis during follow-up increased with increasing PSA levels. The OR for a future prostate cancer diagnosis with a PSA level at baseline of more than 10 ng/mL was 239.5 compared to the reference PSA level of less than 1 ng/mL (Table 3). In paper II, the OR for a prostate cancer diagnosis during follow-up increased with increasing number of risk alleles and the OR was further modified by the PSA level. The evaluated risk alleles performed better in predicting low risk prostate cancer compared to high risk prostate cancer and a future prostate cancer diagnosis in younger men compared to older men (Table 4).

In paper I, the area under the ROC curve for PSA was higher for cases with short lag time compared to cases with long lag time, higher among cases who

0 2 4 6 8 10 12 16 18 20 22 24 26 28 30 32 34 36 38 40 42

No. of risk alleles

Pe rce n t   (%) Controls Cases O dds   ra ti o 0.10 1.00 10

(24)

Table 4 . OR w ith 95% CI o f prost

ate cancer durin

g follow

-up for ca

tegories of the g

enetic risk score

and different leve

ls of PSA. Repr o duced from Johansson et al. , paper II 65 . O verall prosta te cancer Lo w risk prosta te cancer 1

High risk prosta

te cancer 1 A g e at blo od dra w 40 or 50 year s Ag e a t bloo d dra w 60 y e ars Ag e a t diag nosis < 64 Ag e a t diag nosis 64 cases/co ntr o ls 520/988 341/988 179/988 134/259 386/729 237/439 283/549 Categ o ry cases n (% ) contr o ls n (% ) OR (95 % CI) 2 OR (95 % CI) 2 OR (95 % CI) 2 OR (95 % CI) 2 OR (95 % CI) 2 OR (95 % CI) 2 OR (95 % CI) 2

log additive per risk allele

520 988 1.14 (1.11-1. 17) 1.15 (1.11-1. 19) 1.11 (1.06-1. 15) 1.19 (1.11-1. 26) 1.13 (1.09-1. 17) 1.16 (1.11-1. 22) 1.12 (1.08-1. 17) B y le ve ls of PS A a nd nu mber o f risk alleles PS A Risk alleles < 3 ng/mL < 25 23 (4) 169 (17 ) 0.69 (0.43-1. 13) 0.71 (0.40-1. 24) 0.69 (0.28-1. 73) 0.76 (0.31-1. 86) 0.66 (0.37-1. 19) 0.69 (0.32-1. 50) 0.70 (0.37-1. 32) 25 - 31 97 (19) 488 (49 ) 1.00 (refe rence) 1.00 (refe rence) 1.00 (refe rence) 1.00 (refe rence) 1.00 (refe rence) 1.00 (refe rence) 1.00 (refe rence) ≥ 31 83 (16) 199 (20 ) 2.10 (1.50-2. 96) 2.18 (1.49-3. 20) 1.93 (1.04-3. 57) 3.03 (1.71-5. 37) 1.74 (1.13-2. 66) 2.19 (1.30-3. 67) 2.09 (1.33-3. 30) ≥ 3 ng/mL < 25 26 (5) 16 (2) 9.65 (4.92-1 8.9) 6.21 (2.78-1 3.9) 20.2 (8.64-4 7.0) 3.73 (0.59-2 3.7) 10.1 (4.86-2 1.1) 10.5 (3.90-2 8.2) 11.5 (4.25-3 1.0) 25 - 31 147 (28 ) 74 (7) 12.0 (8.30-1 7.4) 9.38 (6.18-1 4.2) 20.9 (12.1-3 6.1) 60.5 (16.7-2 20) 9.20 (6.16-1 3.7) 34.3 (17.0-6 9.0) 7.30 (4.66-1 1.4) ≥ 31 144 (28 ) 42 (4) 20.8 (13.6-3 1.7) 18.4 (11.7-2 9.1) 28.6 (15.7-5 2.2) 37.1 (12.0-1 14) 17.6 (11.1-2 7.9) 44.6 (21.9-9 0.7) 13.0 (7.54-2 2.3)

1) High risk prost

ate cancer define

d as c linical loca l tumour stage T3 or T4, l ymph no de metastasis (N1), bone me tasta sis (M1), Gleaso n score ≥ 8,

WHO grade III, o

r serum levels of PSA at diagnosis > 20 ng/mL; lo

w

risk cancer defined as absence of

all of these factors.

2) Calculated using unconditional l

ogistic regressio n adjusting for ag e and year of rec ruitment.

(25)

Table 5. T he are a under the R O C

curve for differe

nt ages, lag-times, hi

gh/low

risk cancer and for diff

erent risk models in paper I and

II. Table reprod

uced from Holmström et al. , paper I 66 , a nd Johansson et al. , paper II 65 . O verall Lo w risk prosta te cancer 1

High risk prosta

te cancer 1 Ag e a t bloo d dra w < 60 Ag e a t bloo d dra w 60 Lag t ime 2 <2 y e ars Lag t ime 2 <4 y e ars Lag t ime 2 4 y ears Lag t ime 2 >10 y e ars Variables inclu d ed in m odel AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 AU C 3 (95 % CI) 4 Demogra phic 5 0.52 (0.49-0. 55) 0.51 (0.48-0. 55) 0.57 (0.53-0. 62) 0.50 (0.44-0. 56) 0.53 (0.49-0. 57) 0.71 (0.63-0. 79) 0.71 (0.66-0. 76) 0.57 (0.54-0. 60) 0.76 (0.73-0. 80) tPSA 6 0.86 (0.84-0. 88) 0.84 (0.82-0. 86) 0.89 (0.87-0. 92) 0.89 (0.85-0. 92) 0.85 (0.83-0. 87) 0.95 (0.92-0. 98) 0.93 (0.91-0. 96) 0.85 (0.83-0. 87) 0.88 (0.85-0. 91) %fPSA 7 0.78 (0.76-0. 81) 0.75 (0.72-0. 78) 0.86 (0.83-0. 89) 0.78 (0.73-0. 83) 0.78 (0.76-0. 81) 0.85 (0.78-0. 91) 0.86 (0.83-0. 90) 0.78 (0.76-0. 81) 0.85 (0.82-0. 88) tPSA %fPSA 0.86 (0.84-0. 88) 0.85 (0.82-0. 87) 0.90 (0.88-0. 93) 0.89 (0.85-0. 92) 0.86 (0.83-0. 88) 0.95 (0.92-0. 98) 0.93 (0.91-0. 96) 0.85 (0.83-0. 87) 0.88 (0.86-0. 91)

Genetic risk score

0.64 (0.61-0. 67) 0.66 (0.62-0. 69) 0.63 (0.59-0. 68) 0.67 (0.62-0. 73) 0.64 (0.60-0. 67) 0.72 (0.65-0. 79) 0.73 (0.68-0. 78) 0.66 (0.63-0. 69) 0.79 (0.75-0. 82)

tPSA %fPSA Genetic risk score

0.87 (0.85-0. 89) 0.86 (0.84-0. 88) 0.91 (0.88-0. 93) 0.90 (0.87-0. 93) 0.86 (0.84-0. 89) 0.95 (0.92-0. 98) 0.93 (0.91-0. 96) 0.86 (0.85-0. 88) 0.89 (0.87-0. 92)

1) High risk prost

ate cancer defined as cl

inica

l local tumour stage T3 or T4, l

ym ph node metastasis (N1), bone met a stasis (M1), Glea son score ≥ 8, WHO grade III, o r serum levels of PS A at diagnosis > 20 ng/mL; lo w risk c ancer defin

ed as absence of all of these factors. 2) Time

from blood dra

w to diagnosis. 3) Ar ea unde r the R O C curve. 4) C

alculated using unconditional

logistic regression adjusting for ag

e and yea r of re cruitment. 5) A ge at recruitment and year of blood dr a w . 6)

Total PSA. 7) Percent f

ree t

o total PSA.

(26)

26

at the time of recruitment were younger than 59 years compared to cases 59 years or older, and higher for high risk tumours compared to low risk tumours. In paper II, the area under the ROC curve increased by also including percent free to total PSA and the genetic risk score to the model (Table 5).

In both paper I and paper II, the sensitivity, specificity, and the positive and negative likelihood ratios for different cut-offs of PSA were estimated. In paper II, subjects with PSA levels 3 to 10 ng/mL were also reclassified according to the number of risk alleles carried (Table 6). By re-classifying men with PSA between 3 and 10 ng/mL and less than 28 risk alleles as

Table 6. Validity of PSA for prediction of subsequent prostate cancer diagnosis for specific cut-offs of PSA, and subjects with low number of risk alleles subsequently reclassified if having PSA levels of 3-10 ng/mL. Table reproduced from Johansson et al. paper II65.

Cut-off of PSA (ng/mL) Cases 520 Controls 988 Validity by PSA Likelihood ratios

1 Sensitivity Specificity +LR -LR <1 18 434 97% 44% 1.7 0.08 ≥1 502 554 <2 113 742 78% 75% 3.1 0.29 ≥2 407 246 <3 203 856 61% 87% 4.6 0.45 ≥3 317 132 <4 287 909 45% 92% 5.6 0.60 ≥4 233 79 <5 342 936 34% 95% 6.5 0.69 ≥5 178 52 <10 451 977 13% 99% 11.9 0.88 ≥10 69 11

Cut-off for number of risk alleles

Reclassified subjects (PSA 3 - 10 ng/mL and less risk alleles

than cut-off)

Validity after reclassification by

number of risk alleles2Likelihood ratios 1

Cases Control Sensitivity Specificity +LR -LR

<24 9 12 59% 88% 4.9 0.46 <25 17 15 58% 88% 4.9 0.48 <26 29 27 55% 89% 5.2 0.50 <27 45 40 52% 91% 5.6 0.53 <28 66 54 48% 92% 6.1 0.56 <29 94 64 43% 93% 6.2 0.61 <30 113 74 39% 94% 6.7 0.65 <31 139 85 34% 95% 7.2 0.69

1) The positive likelihood ratio was calculated according to [sensitivity/(1-specificity)], and the negative likelihood ratio was calculated as [(1-sensitivity)/specificity].

2) Subjects were first classified using the 3 ng/mL PSA cut-off and subsequently reclassified if having PSA levels in the range of 3-10 ng/mL and less than a specific number of risk alleles.

(27)

27

having negative test the specificity increased from 87% to 92% with a concomitant decrease in sensitivity from 61% to 48%, in comparison to using a PSA cut-off of 3 ng/mL alone. This strategy gave the same specificity as using PSA alone with a threshold of 4 ng/mL, but with a small gain in sensitivity from 45% to 48%. At a fixed specificity of 95%, a sensitivity of 38% was attained by adding percent free to total PSA to PSA. By further adding the genetic risk score to measures of PSA and percent free to total PSA a sensitivity of 44% was achieved.

4.1.2 Discussion

The findings in paper I are in line with data from previous studies evaluating the validity of PSA for early detection of prostate cancer, despite differences in lag time, age at recruitment, and tumour characteristics22-27. Taken together, these results confirm that PSA is a robust biomarker in predicting a future prostate cancer diagnosis. However, when evaluating the validity of PSA as a screening test high standards must be achieved. A specificity of at least 95% and a sensitivity of at least 50% have been suggested for screening tests for diseases with low prevalence67,68.

In paper I, a PSA cut-off of 5.0 ng/mL was required to reach a specificity of 95%, and the sensitivity was 33% at the same PSA cut-off. In order to make a more thorough examination of the classificatory power of a single PSA measure, besides evaluating specificity and sensitivity at different PSA cut-offs, we utilized likelihood ratio estimates. The accuracy in predicting future disease is essential when evaluating the usefulness of a screening test and this can be evaluated by use of likelihood ratios while adapting for various prior risk of disease. A combination of risk factors can be assembled in a risk estimate rather than using a single cut-off28. A positive likelihood ratio above 10 for a diagnostic test is considered to be a strong evidence to “rule in” disease whereas a negative likelihood ratio below 0.1 is considered sufficient evidence to “rule out” disease69. In paper I, no single PSA cut-off resulted in likelihood ratios close to the values proposed for a screening test.

The difficulties in finding a PSA cut-off resulting in a sufficiently high specificity, simultaneously with a reasonably high sensitivity, i.e. above 50%, can be graphically illustrated by the large overlap in the distribution of PSA levels in cases and controls in Figure 3. Furthermore, when analysing common genetic variants associated with prostate cancer, one has to keep in mind that the distribution of the genetic variants in cases and controls show a similar pattern as the distribution of PSA levels, i.e. a large overlap can bee seen (Figure 4). In paper II, no combination of PSA, percent free to total

(28)

28

PSA, and the genetic risk score resulted in likelihood ratios close to the values proposed for a screening test.

A screening test with insufficient specificity and sensitivity might lead to overdiagnosis and overtreatment of prostate cancer, i.e. the identification of prostate cancer that would never have caused symptoms in the patient's lifetime, leading to unnecessary treatment and associated adverse effects70. Despite the identified weaknesses with PSA as a screening test, several randomised controlled trials on prostate cancer screening have shown a decrease in prostate cancer mortality among men who underwent regular PSA testing. In the European Randomised Study of Screening for Prostate Cancer (ERSPC), including 162,243 men aged 55 to 69 years, men attending the first screening round had a rate ratio for death from prostate cancer of 0.73 (95% CI, 0.56 to 0.90) after a median follow-up of 9 years, compared to non-screened men71. In a recent publication from the Göteborg randomised prostate cancer screening trial, with a median follow-up of 14 years, men attending screening had a rate ratio for death from prostate cancer of 0.44 (95% CI, 0.28 to 0.68) compared with the non-screened men72. In the Göteborg trial, 293 men had to be invited to screening and 12 additional men had to be diagnosed with prostate cancer in order to prevent one prostate cancer death. The extent of overdiagnosis is exemplified by the number needed to treat, i.e. the number of additional men needed to be diagnosed and treated with prostate cancer in order to prevent one death from prostate cancer.

The shortcoming of PSA to attain a sufficiently high sensitivity with a concomitant specificity of at least 95% might not be crucial when evaluating PSA for prostate cancer screening. Since screening involves repeated PSA testing with a certain interval, e.g. every second year in the Göteborg screening trial, it might be possible to overcome a low sensitivity by repeated testing. Individuals with prostate cancer have a higher pace in the increase in PSA (i.e. velocity) compared to individuals without prostate cancer73. If the screening intervals are to wide, the amount of interval cancers will increase. Hence, in order to balance the lack of sensitivity and to minimize the amount of interval cancers, individualized screening intervals might be utilized and it has clearly been shown that the PSA levels are highly suggestive of the future risk of a prostate cancer74.

The present limitations with PSA as a screening test, the magnitude of overdiagnosis, and the uncertainties of the cost-effectiveness in PSA-based prostate cancer screening, has led to hesitation about the introduction of

(29)

29

population-based prostate cancer screening75. Most recommendations today on PSA testing, suggests a strategy with informed decision making were men is provided with balanced information on the benefits and harms with early prostate cancer detection before they undergo PSA testing76,77. There is accumulating evidence that supports a strategy with an initial PSA test in all men at a certain age, e.g. men aged 60 years, with a PSA screening follow-up strategy stratified by the PSA level, i.e. prostate cancer risk category. Several studies have suggested that men with PSA levels below average in their specific age category have a low risk of a future prostate cancer that possess threat to life66,78. Such a strategy could be adopted in order to increase the cost-effectiveness of PSA screening and decrease the numbers needed to screen.

Our findings in paper II are in line with the conclusion by Ioannidis et al, i.e. the hitherto identified inherited genetic variants will not give important additional information in stratification of the risk for a future high-risk prostate cancer diagnosis79. This is due to the fact that the present SNPs have been identified mainly among PSA detected low-risk prostate cancers. Future genome-wide association studies in high-risk prostate cancer might find SNPs who adds more information to the individual risk management80.

4.2 Uptake of PSA testing – Paper III 4.2.1 Results

Figure 5 depicts the observed (dotted line) and predicted prostate cancer incidence in absence of PSA testing in men aged 55 to 69 years in Swedish counties 1980 to 2007.The predicted incidence was based on the years 1980 to 1990 using a Poisson model. The estimated annual proportion (with 95% CI) of men aged 55 to 69 years in Swedish counties who had a PSA test using the multiplicative model can be seen in Figure 6, and Figure 7 depicts the same data based on the multiplicative model for the whole country. The estimated cumulated annual proportion of Swedish men aged 55 to 69 years who had a PSA test using the multiplicative and additive models can be seen in Figure 8.

4.2.2 Discussion

The findings in paper III, with a cumulated PSA uptake between 56 and 59% depending on the model used (multiplicative or additive), is at the same level as the proportion of PSA tested men in the control group in the Prostate, Lung, Colorectal and Ovarian (PLCO) screening trial in the sixth year of follow-up82. However, there were considerable regional differences in the PSA uptake in Sweden as illustrated in Figure 6. In a recent publication from

(30)

30

Bratt et al the proportion of Swedish men aged 50 to 75 years having a PSA test between 2000 and 2007 was estimated to at least one-third of all men83.

Figure 5. Observed (dotted lines) and predicted (using a log-linear Poisson model) prostate cancer incidence. G&B=Gothenburg and Bohus county. *) University hospital within the county. Figure reproduced from Jonsson et al.,paper III81.

Figure 6. Estimated annual proportion of PSA testing in Swedish counties. G&B = Gothenburg and Bohus county. *) University hospital within the county. Figure reproduced from Jonsson et al.,paper III81.

Year In ci de nce 200 400 600 800 1985 1995 2005 Älvsborg Blekinge 1985 1995 2005 Dalarna Gävleborg 1985 1995 2005 G & B * Gotland Halland Jämtland Jönköping Kalmar Kristianstad

200 400 600 800 Kronoberg 200 400 600 800

Malmöhus * Norrbotten Örebro * Östergötland Skaraborg Södermanland, Stockholm * 1985 1995 2005 Uppsala * Värmland 1985 1995 2005 Västerbotten * Västernorrland 1985 1995 2005 200 400 600 800 Västmanland Year In ci de nce 200 400 600 800 1985 1995 2005 Älvsborg Blekinge 1985 1995 2005 Dalarna Gävleborg 1985 1995 2005 G & B * Gotland Halland Jämtland Jönköping Kalmar Kristianstad

200 400 600 800 Kronoberg 200 400 600 800

Malmöhus * Norrbotten Örebro * Östergötland Skaraborg Södermanland, Stockholm * 1985 1995 2005 Uppsala * Värmland 1985 1995 2005 Västerbotten * Västernorrland 1985 1995 2005 200 400 600 800 Västmanland Year P ropo rti o n 0.0 0.1 0.2 1985 1995 2005 Älvsborg Blekinge 1985 1995 2005 Dalarna Gävleborg 1985 1995 2005 G & B * Gotland Halland Jämtland Jönköping Kalmar Kristianstad

0.0 0.1 0.2 Kronoberg 0.0 0.1 0.2

Malmöhus * Norrbotten Örebro * Östergötland Skaraborg Södermanland, Stockholm * 1985 1995 2005 Uppsala * Värmland 1985 1995 2005 Västerbotten * Västernorrland 1985 1995 2005 0.0 0.1 0.2 Västmanland Year 0.0 0.1 0.2 1985 1995 2005 Älvsborg Blekinge 1985 1995 2005 Dalarna Gävleborg 1985 1995 2005 G & B * Gotland Halland Jämtland Jönköping Kalmar Kristianstad

0.0 0.1 0.2 Kronoberg 0.0 0.1 0.2

Malmöhus * Norrbotten Örebro * Östergötland Skaraborg Södermanland, Stockholm * 1985 1995 2005 Uppsala * Värmland 1985 1995 2005 Västerbotten * Västernorrland 1985 1995 2005 0.0 0.1 0.2 Västmanland Year 0.0 0.1 0.2 1985 1995 2005 Älvsborg Blekinge 1985 1995 2005 Dalarna Gävleborg 1985 1995 2005 G & B * Gotland Halland Jämtland Jönköping Kalmar Kristianstad

0.0 0.1 0.2 Kronoberg 0.0 0.1 0.2

Malmöhus * Norrbotten Örebro * Östergötland Skaraborg Södermanland, Stockholm * 1985 1995 2005 Uppsala * Värmland 1985 1995 2005 Västerbotten * Västernorrland 1985 1995 2005 0.0 0.1 0.2 Västmanland

(31)

31

Figure 7. Estimated annual proportion of PSA testing in Sweden. Figure reproduced from Jonsson et al.,paper III81.

Figure 8. Cumulated annual proportion of Swedish men aged 55 to 69 years who had a PSA test. Figure reproduced from Jonsson et al.,paper III81.

Data from the Göteborg randomised population-based screening trial utilized for analysis in paper III covered 4 screening visits64. Recently results from 7 screening visits were published from the Göteborg trial72. It is possible that the estimations of the uptake of PSA testing in Sweden could

1990 1995 2000 2005 -0 .0 5 0 .0 0 0 .0 5 0. 10 0. 15 0. 20 Year P ro por tion Multiplicative model 95% CI 1990 1995 2000 2005 0. 0 0 .2 0. 4 0 .6 0. 8 1 .0 Year C u m u la tive p ro por tion Additive model Multiplicative model

(32)

32

have been more accurate if data from 7 screening visits in the Göteborg trial had been utilized. However, the data in paper III was analysed before the publication of the latest results from the Göteborg trial.

In paper III estimations were made for the annual uptake of PSA testing and the cumulated uptake of PSA testing. The assumptions made in the estimations introduce a degree of uncertainty since the actual PSA testing pattern is unknown. Therefore the cumulated number has to be interpreted with caution. However, the exact number of PSA tests might not be of crucial importance but a confirmation of the estimated levels could be of interest for future studies. More importantly the PSA uptake varies considerably between counties. What possible effect this may have on the incidence of metastatic disease and on the prostate cancer specific mortality is unknown. Since both the ERSPC and the Göteborg screening trials have shown a reduction in the prostate cancer specific mortality among those men who underwent regular PSA testing71,72, it is plausible that high PSA uptake counties have a lower prostate cancer specific mortality compared to low PSA uptake counties. However previous studies on a population-based level comparing populations with high versus low uptake of PSA testing have not shown a difference in prostate cancer mortality84,85.

4.3 Primary versus deferred radical prostatectomy –

Paper IV

4.3.1 Results

Characteristics at baseline of men who underwent primary and deferred radical prostatectomy are presented in Table 7. Half of the men that ended surveillance did so due to PSA progression, almost one out of ten due to other signs of progression, and slightly more than two out of five ended surveillance due to other causes. Table 8 depicts the pathology features in specimens from primary and deferred radical prostatectomy, and Table 9 show a comparison of the pathology features in subgroups of different time periods, and Table 10 show a comparison between different geographical areas.

Among those men who underwent primary radical prostatectomy, 16 men (0.7%) had died of prostate cancer, and 145 men (6.2%) had died of other causes at a median follow-up time of 8.2 years after date of diagnosis. Among those men who underwent deferred radical prostatectomy, 2 men (0.9%) had died of prostate cancer, and 12 men (5.4%) had died of other causes after the same follow-up time (Figure 9).

(33)

33

Table 7. Base line characteristics of men who underwent primary and deferred radical prostatectomy. Table reproduced from Holmström et al., paper IV86.

Primary RP1 n=2,344 Deferred RP1 n=222 P Age at diagnosis, years < 0.05 Mean (SD2, range) (5.3, 41 to 70) 61.1 (4.5, 50 to 70) 61.9 <60 880 (38%) 65 (29%) 60 to 64 758 (32%) 91 (41%) 65 to 70 706 (30%) 66 (30%) PSA ng/mL < 0.001 Mean (SD2) 7.8 (3.8) 6.7 (3.4) 0 to 4 262 (11%) 43 (19%) 4 to 10 1484 (63%) 147 (66%) 10 to 20 598 (26%) 32 (14%) cT3 stage < 0.001 T1a 37 (2%) 9 (4%) T1b 27 (1%) 5 (2%) T1c 1363 (58%) 167 (75%) T2 917 (39%) 41 (18%) Gleason score on biopsy n.s. 2 to 4 247 (11%) 35 (16%) 5 525 (22%) 50 (23%) 6 1572 (67%) 137 (62%) Region < 0.001 Stockholm-Gotland 273 (12%) 21 (9%) Uppsala-Örebro 419 (18%) 43 (19%) South-East 163 (7%) 1 (<1%) South 565 (24%) 51 (23%) Western 820 (35%) 97 (44%) Northern 104 (4%) 9 (4%) Year of diagnosis n.s. 1997 103 (4%) 15 (7%) 1998 200 (9%) 26 (12%) 1999 307 (13%) 31 (14%) 2000 488 (21%) 38 (17%) 2001 560 (24%) 59 (27%) 2002 686 (29%) 53 (24%)

(34)

34

Table 8. Adverse pathology features in specimens from primary and deferred radical prostatectomy. Table reproduced from Holmström et al., paper IV86.

Adverse pathology feature OR 1 (95 % CI) P Primary RP2 % Deferred RP2 % Extraprostatic extension 589 of 2,172 27 52 of 205 25 1.1 (0.8 to 1.6) 0.56 Margin positive 716 of 2,176 33 51 of 211 24 0.8 (0.5 to 1.1) 0.10 Gleason score ≥7 569 of 2,318 25 80 of 212 38 2.4 (1.7 to 3.2) <0.001 Any out of three 1168 of 2,141 55 116 of 206 56 1.3 (1.0 to 1.8) 0.09 1) Calculated using unconditional logistic regression adjusting for age, serum PSA, Gleason score, and clinical local tumour stage at the year of diagnosis. 2) Radical prostatectomy. Table 9. Comparison of pathology features in prostatectomy specimens between different periods. Table reproduced from Holmström et al. paper IV86.

1997 to 1999 Primary RP1(%) Deferred RP1(%) OR2(95% CI) P

pT3-43 154/557 28 20/69 29 1.5 (0.8-2.6) p=0.22

Gleason ≥7 120/569 21 23/70 33 2.5 (1.4-4.6) p<0.01

Margin positive 189/572 33 24/69 35 1.4 (0.8-2.5) p=0.20

One out of three 295/549 54 40/68 59 1.6 (0.9-2.8) p=0.08

pT24 margin positive 71/531 13 9/66 14 1.0 (0.5-2.2) p=0.95

2000 to 2002

pT3-43 435/1615 27 32/136 24 1.0 (0.6-1.5) p=0.89

Gleason ≥7 428/1677 26 57/141 40 2.4 (1.7-3.5) p<0.001

Margin positive 527/1604 33 27/142 19 0.5 (0.4-0.8) p<0.01

One out of three 873/1592 55 76/138 55 1.2 (0.8-1.7) p=0.31

pT24 margin positive 216/1525 14 11/134 8 0.6 (0.3-1.1) p=0.09

1) Radical prostatectomy. 2) Calculated using unconditional logistic regression adjusting for age, serum PSA, Gleason score, and clinical local tumour at the year of diagnosis. 3) Pathological tumour stage 3-4. 4) Pathological tumour stage 2.

Table 10. Comparison of pathology features in prostatectomy specimens between different geographical areas. Table reproduced from Holmström et al. paper IV86.

Western Sweden Primary RP1 (%) Deferred RP1 (%) OR2 (95% CI) P

pT3-43 184/697 26 24/91 26 1.3 (0.8-2.2) p=0.35

Gleason ≥7 164/785 21 37/97 38 3.1 (1.9-5.0) p<0.001

Margin positive 209/747 28 19/94 20 0.9 (0.5-1.5) p=0.65

One out of three 376/712 53 54/92 59 1.8 (1.1-2.8) p<0.05 pT24 margin positive 72/648 11 8/88 9 0.9 (0.4-1.9) p=0.70 Excluding western Sweden pT3-43 405/1475 28 28/114 25 1.0 (0.6-1.6) p=0.98 Gleason ≥7 384/1461 26 43/114 38 2.1 (1.4-3.1) p<0.01 Margin positive 507/1429 36 32/117 27 0.8 (0.5-1.2) p=0.21 One out of three 792/1429 55 62/114 54 1.1 (0.7-1.6) p=0.72 pT24 margin positive 215/1408 15 12/112 11 0.7 (0.4-1.2) p=0.20

1) Radical prostatectomy. 2) Calculated using unconditional logistic regression adjusting for age, serum PSA, Gleason score, and clinical local tumour stage at the year of diagnosis. 3) Pathological tumour stage 3-4. 4) Pathological tumour stage 2.

(35)

35

Figure 9. Cumulative mortality due to prostate cancer and due to other causes among men who underwent primary or deferred radical prostatectomy (RP). Figure reproduced from Holmström et al.,paper IV86.

4.3.2 Discussion

The findings in paper IV are in line with previous studies on deferred radical prostatectomy87,88. In a series from the Göteborg part of the ERSPC trial, there were no difference between primary and deferred radical prostatectomy in the frequency of Gleason score more than 6, extraprostatic extension, positive surgical margins, tumour volume on prostatectomy specimens, or for biochemical progression rates after a mean follow-up time of 5.7 years89. In a small US series, the proportion of non-curable disease, defined as less than 75% chance of remaining free from biochemical recurrence at ten years after surgery, after primary and deferred radical prostatectomy was 16% and 23%, respectively88. To date, our study is the only study that has reported on prostate cancer specific mortality for primary versus deferred radical prostatectomy86.

Several studies on active surveillance have shown favourable outcomes after intermediate term follow-up. In a study by Klotz et al with 450 men on active surveillance with selective delayed treatment in case of signs of progression, the 10-year cause specific survival was 97%, and 70% of the men remained on active surveillance after 10 years90. In a study of 616 men treated

(36)

36

expectantly in the ERSPC trial, the calculated 10-year prostate cancer specific survival was 100% and the calculated 10-year treatment-free survival was 43%91. In a study from Carter et al the treatment-free survival was 59% after a median follow-up time of 3.4 years, and 20% (10 of 49) of the men who underwent radical prostatectomy had non-curable disease (defined as above) in the pathology findings88. In a study by Soloway et al, 230 men on active surveillance were followed for a mean time of 44 months. In total 32 men (14%) were subjected to curative treatment after a mean follow-up of 33 months, and for those 12 men who underwent deferred radical prostatectomy, the pathological stage was similar to those men with low risk prostate cancer who underwent primary radical prostatectomy92.

Men that are eligible for active surveillance, i.e. with local clinical tumour stage T1c, Gleason score ≤6 and PSA <10ng/mL, have a low risk of death from prostate cancer. In a recent publication from Eggener et al, the 15 to 20-year prostate cancer specific mortality was negligible since only 3 of 9,557 men who underwent radical prostatectomy with organ confined disease and Gleason score 6 or less had died of prostate cancer93. In a nation-wide Swedish study by Stattin et al the calculated cumulative 10-year prostate cancer specific mortality among men with low risk disease was 2.4% for surveillance and 0.7% among men treated with curative intent (i.e. radical prostatectomy or radiation therapy)52. In the study by Klotz et al, the hazard ratio for death due to competing causes was 33 for men 70 years or older and for men younger than 70 years the corresponding hazard ratio was 990. In a study by Lu-Yao et al of men with a median age of 78 years at diagnosis of a localised prostate cancer managed conservatively, the 10-year risk of dying from competing causes was 60%, 57% and 57% in well, moderately and poorly differentiated prostate cancer94. Hence current evidence supports that men diagnosed with a localised low risk prostate cancer have a low risk of dying from prostate cancer. However, longer follow-up is needed in order to fully elucidate the long-term safety in active surveillance and deferred curative treatment since the risk of dying from prostate cancer in low risk disease is low even on a 20-year basis95,96.

One of the key questions in active surveillance is how to monitor patients safely. In the study by Klotz et al a PSA doubling time of 3 years or less was strongly associated with PSA progression after definite treatment90. In a study by Khatami et al nearly half of men with a PSA doubling time of two years or less had progressed after three years of follow-up while none of the men with a PSA doubling time more than four years had progressed87. These findings indicate that PSA kinetics might be of value in the follow-up

(37)

37

procedure in active surveillance protocols. Most protocols also include repeat biopsies in order to evaluate histologic progression. One important implication in the monitoring of low risk prostate cancer is the risk of undersampling as well as oversampling. Among men who underwent primary radical prostatectomy in the study by Stattin et al52, 25% were upgraded from a Gleason score 6 on core biopsies to a Gleason score more than 6 on the prostatectomy specimens, and 17% were downgraded from a Gleason score 7 on core biopsies to a Gleason score 6 on the prostatectomy specimens (Table 11). The fact that most progression during active surveillance occur one to two years after diagnosis suggests undersampling of a more aggressive tumour rather than progression of a low risk tumour97. Current data suggests that patients diagnosed with a localised low risk prostate cancer have a low risk of progression to incurability if monitored closely.

Table 11. Undergrading and overgrading on core biopsies compared to surgical specimens.

Gleason biopsy Gleason specimen

2-6 (%) 7 (%) 8-10 (%) Total

2-6 1,749 (75) 515 (22) 54 (2) 2,318

7 120 (17) 517 (75) 55 (8) 692

8-10 14 (10) 55 (39) 73 (50) 142

Total 1,833 (60) 1,087 (34) 182 (6) 3,152

Until more solid data are available on the management of localised low risk prostate cancer, treating physicians and their patients have to consider several aspects when deciding on the appropriate approach for localised low risk prostate cancer. The individual choice might be influenced by the information available of the cancer, patient´s expectations, overall health and quality of life issues98.

4.4 Strengths and limitations

4.4.1 Evaluation of validity of PSA and SNPs (Papers I and II)

The external validity is strengthened by the fact that the case mix mirrored that of the source population9, with a standardised incidence rate ratio for prostate cancer of 1.05 (95% CI, 0.96 to 1.16) in the VIP cohort compared to the source population up to 2002. The longitudinal study design enabled us to evaluate the validity of the combination of inherited genetic variants and PSA in predicting a future prostate cancer diagnosis. The true level of PSA uptake in the source population is uncertain and might comprise the estimations of likelihood ratios since populations with higher uptake of PSA testing will give less favourable estimations since they rely on a larger proportion of low risk tumours. The study was hampered by its limited

References

Related documents

As emerging patterns of social and cultural changes new urban housing settlements the so called gated communities can be defined as examples of a residential site of

Дерріда в ролі означення джерела міжтекстової взаємодії у творі використовує формулу «мова, що почила як знак-синал»: «Саме почивши

The default configuration (the one used when using the admin client and specifying the default configuration, or when using the command line tool or API and not supplying a

In line with these priors, our empirical analysis first strongly supports that acquisitions of developed-country firms by EM-MNEs tend to involve relatively strong efficiency

The LES framework supports many forms of deniable authentication [9] through its use of identity-based keys: Alice can create a deniable signature using her secret key SK Alice

Neither district’s placement practices created classrooms with a range of measured student ability and achievement in classrooms sufficiently narrow to be considered

The contemporary consumer freely switches between brick-and-mortar, online and mobile sales channels, expecting that a product will be available any time, any place.. Such

The calculated water levels for MIKE 11 show a delay in the arrival time of the initial surge wave, which was also an observation of the results of the two-dimensional models used