Estimating bias in causes of death ascertainment in the Finnish Randomized Study of Screening for Prostate Cancer.

BACKGROUND: Precise cause of death (CoD) ascertainment is crucial in any cancer screening trial to avoid bias from misclassification due to excessive recording of diagnosed cancer as a CoD in death certificates instead of non-cancer disease that actually caused death. We estimated whether there was bias in CoD determination between screening (SA) and control arms (CA) in a population-based prostate cancer (PCa) screening trial. METHODS: Our trial is the largest component of the European Randomized Study of Screening for Prostate Cancer with more than 80,000 men. Randomly selected deaths in men with PCa (N=442/2568 cases, 17.2%) were reviewed by an independent CoD committee. Median follow-up was 16.8 years in both arms. RESULTS: Overdiagnosis of PCa was present in the SA as the risk ratio for PCa incidence was 1.19 (95% confidence interval (CI) 1.14-1.24). The hazard ratio (HR) for PCa mortality was 0.94 (95%CI 0.82-1.08) in favor of the SA. Agreement with official CoD registry was 94.6% (κ=0.88) in the SA and 95.4% (κ=0.91) in the CA. Altogether 14 PCa deaths were estimated as false-positive in both arms and exclusion of these resulted in HR 0.92 (95% CI 0.80-1.06). CONCLUSIONS: A small differential misclassification bias in ascertainment of CoD was present, most likely due to attribution bias (overdiagnosis in the SA). Maximum precision in CoD ascertainment can only be achieved with independent review of all deaths in the diseased population. However, this is cumbersome and expensive and may provide little benefit compared to random sampling.

Quality-of-life effects of prostate-specific antigen screening.

BACKGROUND: After 11 years of follow-up, the European Randomized Study of Screening for Prostate Cancer (ERSPC) reported a 29% reduction in prostate-cancer mortality among men who underwent screening for prostate-specific antigen (PSA) levels. However, the extent to which harms to quality of life resulting from overdiagnosis and treatment counterbalance this benefit is uncertain. METHODS: On the basis of ERSPC follow-up data, we used Microsimulation Screening Analysis (MISCAN) to predict the number of prostate cancers, treatments, deaths, and quality-adjusted life-years (QALYs) gained after the introduction of PSA screening. Various screening strategies, efficacies, and quality-of-life assumptions were modeled. RESULTS: Per 1000 men of all ages who were followed for their entire life span, we predicted that annual screening of men between the ages of 55 and 69 years would result in nine fewer deaths from prostate cancer (28% reduction), 14 fewer men receiving palliative therapy (35% reduction), and a total of 73 life-years gained (average, 8.4 years per prostate-cancer death avoided). The number of QALYs that were gained was 56 (range, -21 to 97), a reduction of 23% from unadjusted life-years gained. To prevent one prostate-cancer death, 98 men would need to be screened and 5 cancers would need to be detected. Screening of all men between the ages of 55 and 74 would result in more life-years gained (82) but the same number of QALYs (56). CONCLUSIONS: The benefit of PSA screening was diminished by loss of QALYs owing to postdiagnosis long-term effects. Longer follow-up data from both the ERSPC and quality-of-life analyses are essential before universal recommendations regarding screening can be made. (Funded by the Netherlands Organization for Health Research and Development and others.).

Assessment of causes of death in a prostate cancer screening trial.

Accurate assessment of the causes of death is crucial for a conclusive evaluation of the ongoing prostate cancer screening trials. Here, we report the validity of the official causes of death as compared with an independent expert review in the Finnish prostate cancer screening trial. Because nearly 80,000 men were involved, death-cause evaluation was restricted to men diagnosed for prostate cancer. Medical charts were retrieved and the cause of death was assigned by an expert review panel for all deaths among men with prostate cancer during the study period, 1996-2003. The panel decision was compared with both death certificates and the official causes of death as assigned by Statistics Finland. Of a total of 315 deaths, the review panel attributed 127 (41%) to prostate cancer and 184 (59%) to other causes, the corresponding figures in death certificates being 124 (40%) and 187 (60%). Four cases were excluded because of insufficient information. The death-certificate data were in agreement with the panel's assessment in 305 out of 311 cases (overall agreement 97.7%, kappa = 0.95). The overall agreement between the official causes of death and the panel's decision was 97.4% (304/311, kappa = 0.95). The sensitivity of the certificates in identifying prostate cancer deaths was 96.1% (panel as golden standard). Correspondingly, specificity was 98.9%. The official causes of death thus provide an accurate means for evaluating disease-specific mortality in a large population-based prostate-cancer screening trial in Finland.

Second round results of the Finnish population-based prostate cancer screening trial.

PURPOSE: Large randomized trials provide the only valid means of quantifying the benefits and drawbacks of prostate-specific antigen (PSA) screening, but the follow-up of ongoing studies is still too short to allow evaluation of mortality. We report here the intermediate indicators of screening efficacy from the second round of the Finnish trial. EXPERIMENTAL DESIGN: The Finnish trial, with approximately 80,000 men in the target population, is the largest component in the European Randomized Study of Screening for Prostate Cancer. The first round was completed in 1996-1999. Each year 8,000 men 55-67 years of age were randomly assigned to the screening arm, and the rest formed the control arm. Men randomized to the screening arm in 1996 were reinvited 4 years later, in 2000, and PSA was determined. RESULTS: Of the eligible 6415 men, 4407 (69%) eventually participated in the second round of screening. Of the first-round participants, up to 84% (3833 of 4556) attended rescreening. A total of 461 screenees (10.5%) had PSA levels of > or = 4 microg/liter. Altogether, 97 cancers were found, yielding an overall detection rate of 2.2% (97 of 4407). Seventy-nine cases were found among the 3833 second-time screenees (detection rate 2.1%) and 18 in those 574 men (3.1%) who had not participated previously. A PSA of > or = 4 microg/liter, but negative biopsy in the first screening round was associated with an up to 9-fold risk of cancer in rescreening relative to those with lower PSA levels at baseline. Ninety-one (94%) of all of the detected cancers were clinically localized. CONCLUSIONS: As surrogate measures of an effective screening program, both compliance as well as the overall and advanced prostate cancer detection rates remained acceptable. Men defined as screen-positive but with a negative confirmation of cancer at prevalence screen formed a high-risk group at rescreening.

Tumor characteristics in a population-based prostate cancer screening trial with prostate-specific antigen.

PURPOSE: Early diagnosis of prostate cancer is a necessary, but not sufficient, prerequisite for an effective screening program aiming at mortality reduction. We compared tumor characteristics between the screening and control arms in the Finnish population-based screening trial. EXPERIMENTAL DESIGN: The Finnish trial is the largest component in the European Randomized Study of Screening for Prostate Cancer. A total of 24,000 men aged 55-67 years were randomized to the screening arm, whereas 35,973 men formed the control arm during the first three screening years. At the time of invitation, 22,732 men were eligible for screening, and 15,685 (69%) participated. A prostate-specific antigen (PSA) concentration of > or ==" BORDER="0">4 micro g/liter was defined as a screening-positive finding. RESULTS: The detection rate among screenees was 2.4% (377 of 15,685), whereas 0.6% (40 of 7,047) of nonparticipants in the screening arm and 0.3% (112 of 35,973) of the controls were diagnosed with prostate cancer during the first postrandomization year in the absence of screening. In the screening arm, 82% of the cancers were clinically organ confined compared with 65% in the control arm. Yet, both the absolute number and cumulative incidence of advanced cancer were higher in the screening arm. No differences were seen in the WHO grade distribution between the study groups. The median PSA was substantially lower among screen-detected cases (7.1 micro g/liter) than among nonattenders (15.7 micro g/liter) and controls (13.2 micro g/liter). CONCLUSIONS: Our findings on intermediate indicators of PSA screening provide encouraging, yet inconclusive evidence for eventual mortality reduction.

Family history and prostate cancer screening with prostate-specific antigen.

PURPOSE: Early detection of prostate cancer has been recommended for men with affected first-degree relatives despite the lack of evidence for mortality reduction. We therefore evaluated the impact of family history in the Finnish prostate cancer screening trial. PATIENTS AND METHODS: Approximately 80,000 men were identified from the population register for the first screening round. Of the 32,000 men randomized to the screening arm, 30,403 were eligible at the time of invitation. A blood sample was drawn from the participants (n = 20,716), and serum prostate-specific antigen (PSA) was determined. Men with a PSA level > or = 4.0 ng/mL were referred for prostate biopsy. Information on family history was obtained through a self-administered questionnaire at baseline. RESULTS: A total of 964 (5%) of the 20,716 screening participants had a positive family history, and 105 (11%) were screening-positive. Twenty-nine tumors were diagnosed, corresponding to a detection rate of 3.0% (29 of 964) and a positive predictive value of 28% (29 of 105). Of the 19,347 men without a family history, 1,487 (8%) had a PSA level > or = 4.0 ng/mL. The detection rate was 2.4% (462 of 19,347) and the positive predictive value was 31% (462 of 1,487). The risk associated with a positive family history was not substantially increased (rate ratio, 1.3; 95% confidence interval, 0.9 to 1.8). The results were not affected by the age of the screenee or age at diagnosis of the affected relative. The program sensitivity was 6% (29 of 491) (ie, selective screening policy would have missed 94% of cancers in the population). No differences were seen in the characteristics of screen-detected cancers by family history. CONCLUSION: Our findings provide no support for selective screening among men with affected relatives.