SCREENING WITH PROSTATE SPECIFIC ANTIGEN AND METASTATIC
PROSTATE CANCER RISK: A POPULATION BASED CASE-CONTROL
STUDY
JACEK A. KOPEC,*
,† VIVEK GOEL, PETER S. BUNTING, JAN NEUMAN, ERIC C. SAYRE,
PADRAIG WARDE, PETER LEVERS
ANDNEIL FLESHNER‡
From the Department of Health Care and Epidemiology, University of British Columbia (JAK), Arthritis Research Centre of Canada (ECS), Vancouver, Department of Health Policy, Management and Evaluation (VG), and Department of Radiation Oncology (PW), University of Toronto, Division of Urology, Sunnybrook and Women’s College Health Sciences Centre (NF), Leadership Sinai Centre for Diabetes, Mt. Sinai Hospital (JN), Toronto, Department of Laboratory Medicine, The Ottawa Hospital and University of Ottawa, Ottawa
(PSB), Department of Surgery, Southlake Regional Health Centre (PL), Newmarket, Ontario
ABSTRACT
Purpose: Screening of asymptomatic men with prostate specific antigen (PSA) remains a
controversial issue. There is limited evidence that screening is effective in reducing mortality
from prostate cancer. In the current study we determined if screening with PSA reduces the risk
of metastatic prostate cancer.
Materials and Methods: We conducted a population based case-control study among the
residents of Metropolitan Toronto and 5 surrounding counties in Ontario, Canada. Data were
obtained from 236 cases of metastatic prostate cancer and 462 controls randomly sampled from
the source population and frequency matched to cases for age and area of residence. History of
PSA testing, digital rectal examination, symptoms and other data were obtained from medical
records and a self-administered questionnaire. The association between PSA screening and
metastatic prostate cancer was measured by the Mantel-Haenszel odds ratio stratified by
exposure observation time and other potential confounding factors.
Results: In asymptomatic men, the frequency of PSA screening as determined from medical
records was significantly lower among the cases compared with the controls (odds ratio 0.65, 95%
confidence interval 0.45 to 0.93). The odds ratio was 0.52 (0.28 to 0.98) in men 45 to 59 years old
and 0.67 (0.41 to 1.09) in those 60 to 84 years old.
Conclusions: In this case-control study screening of asymptomatic men with PSA was
associ-ated with a significantly reduced risk of metastatic prostate cancer. The results need to be
confirmed in randomized controlled trials.
KEYWORDS: Prostatic neoplasms, diagnosis, prostate-specific antigen, case-control studies
Prostate cancer is the most frequently diagnosed cancer in North American men and the second leading cause of death due to cancer in men.1The epidemiology of prostate cancer has been changed by the introduction of the prostate specific antigen (PSA) in the late 1980s.2The PSA test is used as a screening test, a diagnostic tool and a biological marker to follow the progress of disease in patients with prostate can-cer. Because of widespread use of PSA for opportunistic screening, the incidence of prostate cancer in the United States and many other countries increased dramatically in the early 1990s.3There has also been a shift toward earlier stages at diagnosis.2, 3
Whether regular PSA measurements should be performed in asymptomatic men remains a controversial issue. There is
little evidence to date that screening with PSA is effective in reducing mortality from prostate cancer. A recent case-control study in the US showed a benefit from combined digital rectal examination (DRE) and PSA screening but the data were insufficient to assess the effect of PSA alone.4 Several ecological analyses of population data from the United States and other countries failed to demonstrate a correlation between screening and death rates from prostate cancer5⫺8, while a study in Austria found a significant asso-ciation.9A randomized trial of PSA screening was conducted in Quebec but the results are difficult to interpret because of a low screening rate in the intervention arm.10 Two major randomized trials of screening effectiveness are in progress, 1 in the United States and 1 in Europe, but their results will not be available for several years.11
Given the current controversy regarding the effectiveness of PSA, we conducted a population based case-control study to determine if screening asymptomatic men with PSA re-duces the risk of metastatic prostate cancer. Metastatic pros-tate cancer has a poor prognosis and death from prospros-tate cancer is almost always caused by distant metastases. If screening is effective, we would expect a reduction in meta-static prostate cancer risk before a reduction in mortality. Also, the effect on metastatic prostate cancer should not be confounded by recent improvements in the management of advanced disease.
Submitted for publication October 28, 2004.
Study received University of Toronto Ethics Review Board and participating hospitals ethics review board approval.
Supported by a grant from the National Cancer Institute of Can-ada.
* Correspondence: Department of Health Care and Epidemiology, University of British Columbia, Arthritis Research Centre of Can-ada, 895 West 10th Ave., Vancouver, British Columbia, V5Z 1L7 Canada (telephone: 604-871-4588; FAX: 604-879-3791; e-mail: [email protected]).
† Supported by a National Health Scholar Award from Health Canada.
‡ Financial interest and/or other relationship with BioAdvantex Pharma, Chiron, Aventis, AstraZeneca and Novartis.
See Editorial on page 413.
THEJOURNAL OFUROLOGY® Printed in U.S.A.
Copyright © 2005 by AMERICANUROLOGICALASSOCIATION DOI: 10.1097/01.ju.0000165153.83698.42
METHODS
Data collection.The study population included residents of
Metropolitan Toronto and 5 surrounding counties in Ontario, Canada. Cases were men who developed metastatic prostate cancer between August 1, 1999 and May 31, 2002. They had to be between 40 and 84 years old when diagnosed with metastatic prostate cancer, diagnosed with prostate cancer on or after January 1, 1990, living in the study area at the time of prostate cancer diagnosis and able to answer a ques-tionnaire in English. We identified new cases of metastatic prostate cancer by monthly searching of computerized lists of prostate cancer patients treated at 2 Regional Cancer Cen-ters serving the study population and by regular monthly contacts by telephone, mail, or email with the urologists and oncologists in the study area who would normally treat men with metastatic prostate cancer. Controls were men without metastatic prostate cancer selected randomly from the mu-nicipal tax records database for the study area. Men diag-nosed with nonmetastatic prostate cancer were eligible. The controls were sampled throughout the period of case accrual and frequency matched to the cases for age (5-year intervals) and area of residence (8 areas). In addition, they had to live in the same area as the cases at the time of prostate cancer diagnosis and be able to answer a questionnaire in English. We selected approximately 2 controls for each case. For the purpose of data collection and analysis, each control was assigned a reference date corresponding to the date of pros-tate cancer diagnosis in a case from the same area and age group.
Self-reported data were obtained on age, ethnicity, marital status, education, income, height, weight, employment his-tory, exposure to certain occupational carcinogens, smoking, alcohol use, history of physical activity, diet, vitamin supple-mentation, vasectomy, sexual activity, family history of can-cer, comorbid conditions, use of health services, urinary symptoms, prostate related tests and procedures and history of PSA testing. We reviewed each participant’s medical records obtained from his family physician and, if appropri-ate, 1 or more specialists. Information was collected on PSA history, genitourinary symptoms, history of benign prostatic hyperplasia (BPH) and treatment, digital rectal and trans-rectal ultrasound examinations, previous prostate biopsies, major surgical procedures, second cancers, and other comor-bidity. The charts were reviewed by experienced and trained data abstractors. In most cases the abstractors were blinded to the case-control status of the subject. We performed an inter-abstractor reliability sub-study on a sample of 88 charts. The reason for each PSA test and the results of digital rectal examinations were determined independently by 2 of the investigators (PSB and JN).
Exposure was defined as having a screening (nondiagnos-tic) PSA test. We excluded all tests performed after the diagnosis of prostate cancer or the corresponding reference date for the controls. We also excluded PSA tests where there was evidence in the chart that the physician ordering the test suspected prostate cancer. Furthermore, we excluded tests performed after a positive or suspicious digital rectal exam-ination for prostate cancer or a transrectal ultrasound, a repeat of a previous PSA test and all tests following a positive or suspicious imaging study (bone scan, computerized tomog-raphy or x-ray) for metastatic cancer. We also excluded PSA tests performed after the subject reported frequent daytime urination, post-void dribbling, hesitancy, incomplete empty-ing of bladder, intermittent stream, urgency, weak stream, dysuria, nocturia, hematuria, hemospermia, other genital or urinary symptoms, or bone pain. Although such symptoms do not necessarily indicate prostate cancer, the exclusion of tests performed in symptomatic men reduced the possibility of erroneously counting diagnostic tests as screening. All subjects signed a consent form for participation in the study.
The study was approved by the University of Toronto Ethics Review Board and by the ethics review boards of the partic-ipating hospitals.
Data analysis.Cases and controls were initially compared
in univariate analyses with respect to sociodemographic vari-ables, prostate cancer risk factors, frequency of urinary symptoms, frequency of DRE and health usage variables. In the analysis of PSA screening it was necessary to control for exposure observation time. We analyzed the frequency of PSA screening among cases and controls in a life table con-structed with 2-year intervals starting in 1990. Subjects whose observation time ended in a given interval without a PSA test (for example, because they developed urinary symp-toms or had a positive DRE) were treated as censored obser-vations. We used the propensity score method to control for other confounding variables.12The propensity score was ob-tained from a multiple logistic regression model. Several variables were associated with case-control status at␣⫽0.10 and were retained in the model, namely age, region, family history of prostate cancer, weight, consumption of butter and doctor visits for health problems. The score was divided into tertiles and used as a stratification variable. As a measure of effect, we calculated the stratified Mantel-Haenszel odds ra-tio with test based confidence limits.13Results were obtained for the full sample and for 2 age groups, 45 to 59 and 60 to 84 years old. We conducted a sensitivity analysis to assess the effect of excluding PSA tests in men who had a positive or suspicious DRE. All analyses were done with SAS®.
RESULTS
Data were obtained for 236 cases and 462 controls. The average age at prostate cancer diagnosis among the cases was 68.2 years and the average age at metastatic prostate cancer diagnosis was 71.5 years. Distant metastases were detected by bone scan in 72.1% of the cases. Other methods included computerized tomography (12.7%), x-ray (4.6%), magnetic resonance imaging (4.1%), bone biopsy (4.1%) and other techniques (2.4%).
In univariate analysis, no significant differences between the cases and controls were seen with respect to socioeco-nomic status, immigrant status, most dietary factors, recre-ational and work related activities, frequency of BPH, or frequency of DREs (table 1). Cases were heavier by an aver-age of 5.8 lbs and were more likely to have had a family member with prostate cancer. As shown in the figure cases were less likely to report urinary symptoms, except for the final year before the diagnosis/reference date.
In crude data 58 cases (24.6%) and 126 controls (27.3%) had a screening PSA test (table 2). The overall Mantel-Haenszel odds ratio, adjusted for exposure observation time as well as matching and confounding variables, was 0.65 (95% confidence interval 0.45 to 0.93) in favor of screening. When the data were analyzed according to age, the odds ratio was 0.52 (0.28 to 0.98) for men 45 to 59 years old and 0.67 (0.41 to 1.09) for those 60 to 84 years old.
DISCUSSION
In this population based case-control study, we found that screening of asymptomatic men with PSA reduced their risk of metastatic prostate cancer by 35%. The effect is clinically important, especially considering that the majority of the exposed controls were screened only once. We observed a benefit of screening in men less than 60 years old as well as those 60 years old or older. The association was stronger in younger men, however the study did not have enough power to demonstrate a significant difference in effect according to age.
A number of methodological issues should be considered when interpreting the findings. Cases in this study came from a geographically defined population and controls were
randomly selected from the same source population. Al-though we may have missed some cases treated outside of the Regional Cancer Centers, such cases would have been rare. Response rates to the mailed questionnaire were 69% among the cases and 51% among the controls, ie in the range typi-cally found in mail population surveys in North America. In a study population composed mostly of older men, almost half of them born outside Canada, such response rates seem rea-sonable. Medical records were obtained for 90% of the respon-dents (93% of the cases and 88% of the controls). Because the
subjects were not informed about the study hypothesis, it seems unlikely that their participation would have been linked to PSA screening. Nevertheless, a potential for selec-tion bias is a limitaselec-tion of the study.
Our analysis included 36 cases that were asymptomatic at the time of prostate cancer diagnosis and had a positive PSA test within the preceding 6 months. While some of these cases may not have been captured by our study in the ab-sence of screening, excluding any such cases would have increased the effect in favor of screening.
We controlled for the effect of matching by age and area of residence, as well as confounding by suspected risk or pro-tective factors for prostate cancer and screening with digital rectal examination. We considered the possibility that men who undergo screening may be more health conscious and more likely to engage in other preventive strategies or seek medical care for any problems, and that such behaviors may protect against metastatic prostate cancer. To control for the effect of health care usage behavior, we included visits to doctors as a potential confounding variable. Confounding by unknown factors can never be ruled out in a nonrandomized study. However, such factors would have to be strongly as-sociated with metastatic prostate cancer and PSA screening. Differences between the crude and adjusted odds ratios were primarily due to longer observation times in the cases. The main reason was a later onset of symptoms in the cases, especially in the older age group (see figure, table 2).
Differential misclassification of PSA information in medi-cal records seems unlikely since most of the reviews were done blindly and the abstractors followed detailed instruc-tions. Our reliability study demonstrated excellent agree-ment between the abstractors. We found 96.8% agreeagree-ment for having a PSA, 93.9% for a digital rectal examination and 91.1% for symptoms of prostatism. The level of agreement was also high for the coding of digital rectal examination results (96.6%) and reasons for PSA testing (90.7%).
We defined a screening (nondiagnostic) PSA test as one performed before the onset of symptoms, as recommended in the literature.14 Although prostate related symptoms are usually associated with BPH, similar symptoms may be present in patients with prostate cancer. The dramatic in-crease in the frequency of symptoms just before the diagnosis of prostate cancer suggests that the inclusion of symptomatic men would have resulted in a substantial misclassification of diagnostic tests as screening. We classified PSA tests per-formed after a positive or suspicious DRE as diagnostic rather than screening. It is possible that some of these tests would have been done anyway for screening purposes. In 16 subjects, 15 cases and 1 control, reclassifying such tests as screening would have changed their status from unexposed to exposed. All these tests were done within 3 months of the
TABLE 1. Distributions of selected self-reported variables among cases and controls
Variable Cases (%) Controls (%) Age at diagnosis of prostate Ca:
45–54 11 (4.7) 29 (6.3) 55–64 66 (28.0) 124 (26.8) 65–74 110 (46.6) 226 (48.9) 75–84 49 (20.8) 83 (18.0) Country of birth: Canada 126 (53.4) 254 (55.0) Other 108 (45.8) 207 (44.8) Not stated 2 (0.8) 1 (0.2) Ethnicity: White 185 (78.4) 372 (80.5) Black 6 (2.5) 3 (0.6) Other 43 (18.2) 86 (18.6) Not stated 2 (0.8) 1 (0.2) Marital status: Married or common-law 193 (81.8) 396 (85.7) Single, divorced or widowed 42 (17.8) 65 (14.1)
Not stated 1 (0.4) 1 (0.2)
Education:
Less than high school 76 (32.2) 144 (31.2) High school with or without
trade/technical training
76 (32.2) 147 (31.8)
College or university 80 (33.9) 169 (36.6)
Not stated 4 (1.7) 2 (0.4)
Gross household income:
Less than $30,000 26 (11.0) 58 (12.6) $30,000–49,999 48 (20.3) 92 (19.9) $50,000–100,000 42 (17.8) 114 (24.7) More than $100,000 16 (6.8) 49 (10.6) Not stated 104 (44.1) 149 (32.3) Usual type of activity in daily
work: Sitting 74 (31.4) 161 (34.8) Light 64 (27.1) 128 (27.7) Moderate 50 (21.2) 93 (20.1) Strenuous 47 (19.9) 69 (14.9) Not stated 1 (0.4) 11 (2.4)
Relative diagnosed with pros-tate Ca:* Yes 51 (21.6) 56 (12.1) No 178 (75.4) 401 (86.8) Not stated 7 (3.0) 5 (1.1) Urinary symptoms:*,† Yes 68 (28.8) 181 (39.2) No 161 (68.2) 274 (59.3) Not stated 7 (3.0) 7 (1.5) Diagnosis of BPH: Yes 50 (21.2) 116 (25.1) No 185 (78.4) 345 (74.7) Not stated 1 (0.4) 1 (0.2)
Visits to doctors for routine ex-amination:
Never 28 (11.9) 40 (8.7)
Once in 2–5 yrs 65 (27.5) 108 (23.4) Once/yr or more 139 (58.9) 309 (66.9)
Not stated 4 (1.7) 5 (1.1)
Ever had DRE:
Yes 188 (79.7) 395 (85.5)
No 40 (16.9) 60 (13.0)
Not stated 8 (3.4) 7 (1.5)
* pⱕ0.05.
† Urinary symptoms were defined as being bothered 10 years ago, at least “half the time,” by 1 or more of symptoms of a sensation of not emptying bladder after urination, having to urinate again less than 2 hours after finishing, stopping and starting several times during urination, difficulty postponing urination, weak urinary stream, pushing or straining to begin urination or getting up to urinate during the night.
Cumulative proportion of cases and controls with prostate related symptoms before prostate cancer diagnosis/reference date.
DRE. Additional analyses showed that in about 20% of neg-ative DREs, a screening PSA test followed within 3 months. Applying this proportion, we performed a sensitivity analysis assuming that 3 of the 15 cases with a PSA test following a suspicious/positive DRE may have been done for screening. With this assumption the result would remain virtually un-changed (odds ratio 0.68, 0.48 to 0.97).
Previous studies offered limited evidence for the effective-ness of screening for prostate cancer. A recent case-control study by Weinmann et al produced an odds ratio of 0.70 in favor of screening with DRE and PSA. However, most of the tests were DREs and it was not possible to assess the sepa-rate effect of PSA screening.4In other case-control studies of screening with DRE, Friedman et al15and Richert-Boe et al16 reported nonsignificant results, while Jacobsen et al17found a strong inverse association. In a randomized trial of PSA screening, Labrie et al reported an odds ratio of 0.31 when comparing screened and unscreened men.10 Unfortunately only 23.1% of those randomized to screening were actually tested and an intent to screen analysis did not show a sig-nificant effect. Lu-Yao et al compared prostate cancer mor-tality up to 1997 among Medicare beneficiaries in the Seattle and Connecticut cancer surveillance areas.5Mortality rates were similar in the 2 areas although PSA testing was 5.4 times more common in Seattle. In contrast Bartsch et al demonstrated a significantly greater rate of prostate cancer mortality decrease after the introduction of mass screening with PSA in the province of Tyrol compared with the rest of Austria.9A number of authors used prostate cancer incidence as a surrogate for screening frequency. These studies gener-ally found no association between increased incidence and decreases in mortality.6⫺8However, the ecological design is inherently prone to bias. Other factors, such as improved treatment of prostate cancer and risk factor changes, may contribute to mortality trends.3Furthermore, followup may not have been long enough in some studies.5
CONCLUSIONS
Although the strongest evidence for or against PSA efficacy will come from large randomized trials, the results of such trials are several years away. Moreover, generalizability of the current trials may be limited by their eligibility criteria, self-selection of participants and other design related is-sues.11Our study provides suggestive evidence for the effec-tiveness of screening with PSA as actually practiced in the study area in the 1990s and early 2000s. It is possible that a greater benefit would result from a screening program with regular screening intervals. Because screening for prostate cancer is a complex issue, more studies in different popula-tions and using a variety of approaches are needed. It will be important to further delineate the effect of PSA in men with prostate related symptoms and estimate the optimal fre-quency of screening. While reduction in mortality and
meta-static cancer are important, quality of life issues and men’s preferences need to be carefully evaluated.
Cyrelle Gold reviewed medical records, and Vartouhi Jazmaji, Nicole Riley and Deva Thiruchelvam assisted with database management and data entry.
REFERENCES
1. Weir, H. K., Thun, M. J., Hankey, B. F., Ries, L. A., Howe, H. L., Wingo, P. A. et al: Annual report to the nation on the status of cancer, 1975–2000, featuring the uses of surveillance data for cancer prevention and control. J Natl Cancer Inst,95:1276, 2003
2. Hankey, B. F., Feuer, E. J., Clegg, L. X., Hayes, R. B., Legler, J. M., Prorok, P. C. et al: Cancer surveillance series: interpret-ing trends in prostate cancer—part I: evidence of the effects of screening in recent prostate cancer incidence, mortality, and survival rates. J Natl Cancer Inst,91:1017, 1999
3. Potosky, A. L., Feuer, E. J. and Levin, D. L.: Impact of screening on incidence and mortality of prostate cancer in the United States. Epidemiol Rev,23:181, 2001
4. Weinmann, S., Richert-Boe, K., Glass, A. G. and Weiss, N. S.: Prostate cancer screening and mortality: a case-control study (United States). Cancer Causes Control,15:133, 2004 5. Lu-Yao, G., Albertsen, P. C., Stanford, J. L., Stukel, T. A.,
Walker-Corkery, E. S. and Barry, M. J.: Natural experiment examining impact of aggressive screening and treatment on prostate cancer mortality in two fixed cohorts from Seattle area and Connecticut. BMJ,325:740, 2002
6. Coldman, A. J., Phillips, N. and Pickles, T. A.: Trends in prostate cancer incidence and mortality: an analysis of mortality change by screening intensity. CMAJ,168:31, 2003
7. Perron, L., Moore, L., Bairati, I., Bernard, P. M. and Meyer, F.: PSA screening and prostate cancer mortality. CMAJ,166:586, 2002
8. Threlfall, T. J., English, D. R. and Rouse, I. L.: Prostate cancer in Western Australia: trends in incidence and mortality from 1985 to 1996. Med J Aust,169:21, 1998
9. Bartsch, G., Horninger, W., Klocker, H., Reissigl, A., Oberaigner, W., Schonitzer, D. et al: Prostate cancer mortality after intro-duction of prostate-specific antigen mass screening in the Fed-eral State of Tyrol, Austria. Urology,58:417, 2001
10. Labrie, F., Candas, B., Dupont, A., Cusan, L., Gomez, J.-L., Suburu, R. E. et al: Screening decreases prostate cancer death: first analysis of the 1988 Quebec prospective randomized con-trolled trial. Prostate,38:83, 1999
11. de Koning, H. J., Auvinen, A., Berenguer Sanchez, A., Calais da Silva, F., Ciatto, S., Denis, L. et al: Large-scale randomized prostate cancer screening trials: program performances in the European Randomized Screening for Prostate Cancer trial and the Prostate, Lung, Colorectal and Ovary cancer trial. Int J Cancer,97:237, 2002
12. Joffe, M. M. and Rosenbaum, P. R.: Invited commentary: pro-pensity scores. Am J Epidemiol,150:327, 1999
13. Kahn, H. A. and Sempos, C. T.: Statistical Methods in Epidemi-ology. New York: Oxford University Press Inc., 1989 14. Cronin, K. A., Weed, D. L., Connor, R. J. and Prorok, P. C.:
Case-control studies of cancer screening: theory and practice. J Natl Cancer Inst,90:498, 1998
TABLE 2. Adjusted odds ratios for association between screening with PSA and risk of metastatic prostate cancer in asymptomatic men
Cases Controls Time Adjusted Odds Ratio* 95% Confidence Interval Fully Adjusted Odds Ratio† 95% Confidence Interval Total 236 462
Not PSA screened 178 336 1.00 — 1.00 —
PSA screened 58 126 0.66 0.47, 0.93 0.65 0.45, 0.93
Age 45–59 101 195
Not PSA screened 83 138 1.00 — 1.00 —
PSA screened 18 57 0.51 0.28, 0.93 0.52 0.28, 0.98
Age 60–84 135 267
Not PSA screened 95 198 1.00 — 1.00 —
PSA screened 40 69 0.73 0.46, 1.15 0.67 0.41, 1.09
* Mantel-Haenszel odds ratio stratified by exposure observation time.
† Mantel-Haenszel odds ratio stratified by exposure observation time and tertiles of the propensity score. Several variables were included in the propensity score, namely age, region, family history of prostate cancer, weight, consumption of butter and doctor visits for health problems.
15. Friedman, G. D., Hiatt, R. A., Quesenberry, C. P., Jr. and Selby, J. V.: Case-control study of screening for prostatic cancer by digital rectal examinations. Lancet,337:1526, 1991
16. Richert-Boe, K. E., Humphrey, L. L., Glass, A. G. and Weiss, N. S.: Screening digital rectal examination and prostate cancer mortality: a case-control study. J Med Screen,5:99, 1998 17. Jacobsen, S. J., Bergstralh, E. J., Katusic, S. K., Guess, H. A.,
Darby, C. H., Silverstein, M. D. et al: Screening digital rectal examination and prostate cancer mortality: a population-based case-control study. Urology,52:173, 1998
EDITORIAL COMMENT
Kopec et al report that history of a screening PSA test was 35% less frequent among men in whom metastatic prostate cancer devel-oped than among controls randomly sampled from the general pop-ulation. The apparent reduction in risk is notable in that most of the 27% of controls who had a screening PSA were only screened once. Results from other, primarily nonrandomized studies have been inconsistent.
Since PSA screening does not prevent prostate cancer, the most appropriate case-control design would have compared PSA screening among prostate cancer cases in whom metastases did and did not develop. The validity of the current design depends on whether PSA
use in controls (98% of whom did not have prostate cancer) is con-founded with other health behaviors that would be expected to differ between men with and without prostate cancer. Controlling for con-founding factors with the use of the propensity score may not have adjusted any imbalance in overall health behavior since variables that may correlate with health behavior (smoking, alcohol, income, physical activity) were not in the model. Although none of these variables was individually associated with metastatic prostate can-cer at p⬍0.10, small differences could jointly signify differences in health behavior. Bias could also arise if recall of past urinary symp-toms was underreported by patients with metastatic prostate cancer as events since their diagnosis, treatment and side effects of treat-ment assumed greater prominence. However, it is also possible that the protective effect of screening is somewhat underestimated since PSA usage was low during the early part of the study period. The authors are aware of these potential limitations and recognize the need for a randomized study of mitigate biases in the controls. Despite these potential concerns the study provides further evidence that PSA screening can prevent death from prostate cancer.
Bruce J. Trock
Brady Urological Institute The Johns Hopkins Hospital Baltimore, Maryland