Prostate Cancer Screening Practices and Beliefs
A Longitudinal Physician Survey†
* Partial results of this study were presented in abstract form at the annual meeting of the National Society of General Internal Medicine, May, 2000, Boston, Mass.
* Address correspondence and reprint requests to Dr. Voss: Division of General Medicine, Box 800744, Primary Care Center, University of Virginia Health System, Charlottesville, VA 22908.
OBJECTIVE: To examine primary care physician prostate-specific antigen (PSA) testing and prostate cancer screening beliefs, practices, and trends over time.
DESIGN: Longitudinal physician survey.
SETTING: Community and academic primary care practices in a major East Coast city.
PARTICIPANTS: Physicians with ≥100 patients enrolled in an HMO serving 90,000 patients were surveyed in 1993 and 1998 regarding prostate cancer screening beliefs and practices. In 1993, 176 physicians (76%) completed the survey. In 1998, 76% of the 1993 respondents responded to a second survey. Associations between and changes over time in beliefs and self-reported PSA testing were analyzed.
MEASUREMENTS AND MAIN RESULTS: Physicians reported ordering PSA tests in 73% of health maintenance exams (HMEs) in 1993 and 81% of HMEs in 1998. PSA testing significantly increased between 1993 and 1998 with 43% of physicians reporting increased testing and only 13% reporting reduced testing. Between 1993 and 1998, physician attitudes favoring PSA testing increased although less than half of physicians believed that aggressive early treatment improved patient outcomes. Community versus academic practice location was also associated with PSA testing.
CONCLUSIONS: Physicians reported high and increasing rates of PSA testing from 1993 to 1998 and more favorable attitudes toward PSA testing. Despite conflicting expert recommendations and a lack of consistent, high-quality supporting evidence, PSA screening appears to be increasingly considered a standard of care by practicing physicians. However, beliefs other than improved patient outcomes due to screening may be the primary drivers of increased PSA testing.
Prostate cancer is the most common nonskin cancer and the second leading cause of male cancer deaths among American men. It is estimated that 198,100 new cases of prostate cancer will be diagnosed among American men in 2000.1 The high incidence of this cancer, the aging of the population, the prominence and advocacy of individuals who have suffered it and organizations concerned about it, as well as the tendency for our society to believe that aggressive intervention is best, have led to a surge in interest and use of screening for this disorder over the past decade. Recent evidence suggests that the dramatic fluctuations in prostate cancer incidence observed in the early 1990s may at least partially be attributed to the introduction and widespread use of prostate-specific antigen (PSA) as a screening test.2 However, controversy exists about the value of screening for prostate cancer in asymptomatic men, because there are no completed randomized trials that have demonstrated a reduction in prostate cancer mortality from screening.3
Prostate cancer screening recommendations from professional and governmental institutions have conflicted over the past decade, leading to potential uncertainty among physicians regarding appropriate screening practices. Major specialty and preventive health guideline groups, including the American College of Physicians,4 the U.S. Preventive Services Task Force,5 the American College of Preventive Medicine,6 and the American Academy of Family Practice,7 have stated that the evidence supporting PSA screening is insufficient to recommend its routine use. Although the digital rectal examination (DRE) has been considered by many to be ‘grandfathered’ in as a screening examination, these groups have all cautioned that PSA testing should only be considered a screening option until evidence indicates that its use results in a reduction in prostate cancer–specific morbidity and/or mortality. On the other hand, both the American Cancer Society8 (ACS) and the American Urologic Association9 have been major advocates of the use of PSA screening for prostate cancer over the past decade. The American Cancer Society advocacy softened somewhat in the late 1990s, as evidenced by the 1997 ACS updated prostate cancer screening guideline indicating that patients should be provided with information regarding potential benefits and harms of screening rather than directly recommending PSA testing for all men over age 50 or at high risk of prostate cancer.10 The most recent ACS position (May, 2000) states that all average-risk men aged 50 or older with a life expectancy of at least 10 years should be offered a PSA test annually and be provided information about the benefits and limitations of testing, but does not generally endorse mass screening, i.e., testing outside of an established clinician-patient relationship.11
Despite a paucity of supporting evidence, surveys of physicians conducted in the early to mid 1990s have demonstrated variable but relatively high rates of PSA utilization for prostate cancer screening, consistent with the substantial publicity and advocacy that have occurred. Although there are several published surveys of PSA screening behavior,12–18 we found none that evaluated the change in such behavior over time. We report the results of a longitudinal study that surveyed self-reported screening practices, knowledge, attitudes, and beliefs in a sample of primary care physicians in 1993 and in 1998 to identify determinants of and trends in prostate cancer screening.
The study population was comprised of general internists and family physicians participating in a 90,000-member mixed model health maintenance organization affiliated with a major academic medical center serving metropolitan Washington, D.C. Physicians participated in either a network model composed of community physicians or a group model composed of academic physicians employed by the medical center. The survey asked physicians to report their use of DRE and PSA testing to screen asymptomatic men aged 50 or older. Other survey items asked respondents about their beliefs regarding potential determinants of screening behavior.
In 1993, all general internists and family physicians with more than 100 patients in the health maintenance organization received up to 3 survey mailings. One hundred seventy-six survey responses were received. The overall survey response rate was 76%, with a survey response rate of 73% among network-model physicians and 85% among group-model physicians. In 1998, the initial 176 survey respondents were resurveyed using the original questionnaire. Thirty-five physicians could not be located or had retired from active practice. One hundred eight survey responses were received, yielding an adjusted response rate of 76% (after exclusion of the 35 physicians not surveyed). The survey response rate was 74% among community physicians and 85% among academic physicians in 1998. The final data set consisted of 108 physicians with matched survey responses from both the 1993 and 1998 questionnaires.
Physicians were asked to estimate the percentage of times they ordered PSA tests and the percentage of times they performed DREs on healthy men aged ≥50 presenting for routine health maintenance examinations (HMEs). Respondents selected a percentage ranging from 0% to 100% on a horizontal 10% interval scale for each of these 2 questions. Physicians were categorized as low PSA test orderers (0% to 30%), moderate test orderers (40% to 70%) or high test orderers (80% to 100%), these categories representing natural breakpoints based on examination of the 1993 test-ordering frequency distribution. Physicians indicated whether they agreed with, were uncertain about, or disagreed with specific attitudes or beliefs regarding prostate cancer screening or treatment. To identify physicians most predisposed to PSA testing, we collapsed the 3 levels of response into responses favoring PSA testing versus uncertain or negative PSA testing responses. Demographic information collected on the 1998 survey included age, gender, year of graduation from medical school, specialty, practice type, percentage of time spent in clinical practice, teaching, administration or research, and percentage of managed care patients in each physician's practice. In 1998, physicians were also asked to estimate the percent of PSA tests they ordered based on patient request as the primary testing motivation.
Data were analyzed using SPSS for Windows (version 9.0; SPSS, Inc., Chicago, Ill). X2 tests of association or other nonparametric tests were used to analyze nominal data, and analysis of variance was used to analyze continuous data. A 2-tailed P < .05 was employed as the criterion for statistical significance.
Table 1 illustrates demographic characteristics of the physician group. There was no significant difference in 1993 responses between 1998 survey responders and nonresponders. Wide variation in physician PSA testing practices was observed in 1993 and 1998. Physician responses ranged from never (0%) to always (100%) ordering screening PSA tests for health maintenance examinations. From 1993 to 1998, physicians reported a rise in mean percentage of times they ordered PSA tests from 72.8% to 80.5%, an absolute increase of 7.7% (confidence interval [CI] 2.6% to 12.9%; P = .004), and a relative increase of 10.6%. In a pairwise comparison from 1993 to 1998, 43.3% of physicians reported no change in test ordering, 43.3% reported increased PSA test ordering and 13.4% reported decreased PSA test ordering (Wilcoxon signed ranks P = .001). Among high PSA testers in 1993, 87% continued to report high test ordering in 1998. Among moderate testers in 1993, 83% changed to high test ordering in 1998, 17% remained moderate testers, and none changed to low testers. Among low testers in 1993, 18% reported high test ordering in 1998, 32% reported moderate testing, and 50% continued to report low test ordering (McNemar P = .024) (Table 2). In contrast, physicians reported performing DRE in 94% and 96% of screening ambulatory HMEs in 1993 and 1998, respectively (P = .110).
Table 1. Physician Characteristics
| Male||59 (63)|
| Female||35 (37)|
| Academic||22 (22)|
| Community||79 (78)|
| 30–39||16 (15)|
| 40–49||43 (41)|
| 50–59||34 (32)|
| 60+||13 (12)|
|Years since graduation from medical school|
| 0–10||10 (11)|
| 11–20||39 (42)|
| 21–30||28 (30)|
| 31+||16 (17)|
| General internal medicine||63 (64)|
| Family medicine||26 (27)|
| Other||9 (9)|
|Percent managed care|
| 0–25||(7) 7|
| 26–50||35 (36)|
| 51–75||30 (31)|
| 76–100||25 (26)|
Table 2. PSA Testing Status Concordance from 1993 to 1998
|High testers, n (%)||55 (87.3)||6 (9.5)||2 (3.2)||63 (100)|
|Moderate testers, n (%)||10 (83.3)||2 (16.7)||0||12 (100)|
|Low testers, n (%)||4 (18.2)||7 (31.8)||11 (50)||22 (100)|
Physician attitudes and beliefs regarding prostate cancer screening in 1993 and 1998 are presented in Table 3. Significant increases in agreement with attitudes and beliefs favoring PSA testing were observed from 1993 to 1998. In both 1993 and 1998, PSA test ordering was highly associated with belief that PSA testing represented standard of care practice, that failure to order PSA tests increased potential malpractice liability, and that PSA tests had acceptable operating characteristics for screening purposes. These data are illustrated in Table 4, where the P value indicates the significance of the association between reported PSA testing and agreement with each statement. In the 1998 but not the 1993 survey, we found a significant positive relationship between physician belief in the benefits of aggressive therapy (radical prostatectomy) and PSA testing. It is notable that only 43.6% of physicians in 1998 indicated that they believed that radical prostatectomy reduced prostate cancer–specific mortality. In both 1993 and 1998, an inverse relationship between PSA testing and belief that treatment harms might outweigh treatment benefits was observed. However, more than half of the physicians agreed or were unsure whether prostate cancer treatment harms might outweigh treatment benefits.
Table 3. Physician Attitudes and Beliefs
| Yearly PSA determinations for asymptomatic men age ≥50 are a standard of care.||53||64||.006|
| Providers face malpractice liability if PSA not performed and cancer detected later.||36||53||.003|
| Radical prostatectomy for early prostate cancer reduces prostate cancer–specific mortality.||29.3||43.6||.013|
| PSA has acceptable sensitivity, specificity, and positive predictive value to be used as a screening test.||46.9||59.2||.036|
| PSA should be routinely employed to screen all men beginning at age 50.||53.1||67.3||.013|
| PSA should be routinely employed to screen all high risk men (African-American men, 1st deg relative with prostate cancer) beginning at age 40.||54.2||75||.001|
| Morbidity and mortality of treating prostate cancer may outweigh benefits of treatment.||35.4||46.5||.071|
Table 4. Physician Attitudes and Beliefs versus Self-reported PSA Testing
| Yearly PSA determinations for asymptomatic men age ≥50 are a standard of care.||1993||65.1||25||9.1||<.001|
| Providers face malpractice liability if PSA not performed and cancer detected later.||1993||46||25||13||.017|
| Radical prostatectomy for early prostate cancer reduces prostate cancer–specific mortality.||1993||30.6||25||31.8||.911|
| PSA has acceptable sensitivity, specificity, and positive predictive value to be used as a screening test.||1993||68.9||16.7||4.5||<.001|
| PSA should be routinely employed to screen all men beginning at age 50.||.1993||82.5||16.7||0||<.001|
| PSA should be routinely employed to screen all high risk men (African-American men, 1st deg relative with prostate cancer) beginning at age 40.||1993||58.7||91.7||22.7||<.001|
| Morbidity and mortality of treating prostate cancer may outweigh benefits of treatment.||1993||47.6||16.7||13.6||.006|
Table 5 illustrates the relationships between PSA testing and selected physician characteristics. Age and years since graduation from medical school were associated with PSA testing in 1993 and 1998, and a trend toward greater PSA testing among male physicians was seen in 1998 but did not reach statistical significance. Community versus academic practice model was significantly associated with PSA testing in both 1993 and 1998. In 1998, PSA testing was positively associated with the percentage of time physicians were engaged in clinical practice but was negatively associated with the percentage of time physicians were engaged in nonclinical activities (defined as teaching, research, or administration). There was no association between physician estimates of managed care penetration of their practice and PSA testing.
Table 5. Physician Characteristics and PSA Testing Practices
|Community model, %||1993||95.2||100.0||22.7||<.0001|
|Managed care, %||1998||58.3||58.6||63.9||.72|
|Clinical practice, %||1998||91.9||78.0||53.2||<.0001|
|Nonclinical practice, %||1998||8.2||21.4||37.9||<.0001|
In the 1998 survey, physicians reported that a mean of 31.3% (range 0% to 100%) of PSA tests were ordered primarily due to patient demand. There was a strong inverse association between the frequency of patient demand–based testing and high, moderate, or low overall PSA testing practices. High overall PSA test orderers were patient demand driven 25.8% of the time compared with 43.5% for moderate test orderers and 65.4% for low test orderers (P < .001).
To explore a possible logical contradiction between test ordering and beliefs, we performed post-hoc analyses on the subset of patients failing to endorse the efficacy of early prostate cancer treatment to discover alternative factors associated with high testing. In 1998, 87.5% of high testing physicians but only 14.0% of low to moderate testing physicians felt PSA tests represented a standard of practice (P < .001). Sixty-two percent of high testers but only 4.8% of low to moderate testers felt PSA tests had suitable screening test characteristics (P < .001). High testers were also more likely to believe that the treatment harms did not outweigh treatment benefits, although 59% of even high testers in 1998 did indicate belief that the treatment harms might outweigh the treatment benefits. We observed a trend indicating that high testers were more likely to agree that failure to order PSA tests posed a malpractice liability than were low to moderate testers (53% vs 33%; P = .157). In this subgroup of physicians who did not affirm the efficacy of aggressive treatment of early prostate cancer, high testing physicians were older (mean age 47.5 vs 42.2 years; P = .006), more distant from medical school graduation (20.4 vs 15.0 years; P = .021), and more likely to be men (63.3% vs 26.3%; P = .012). In 1998, 94% of high testing physicians were from community-based practices while 76.2% of low to moderate testers were from academically based practices (P < .0001). Subgroup analysis of 1993 survey data revealed similar response patterns.
Primary care physicians surveyed in this study reported high and increasing rates of PSA testing for asymptomatic men over age fifty. Physicians also demonstrated increasing agreement from 1993 to1998 with attitudes favoring PSA testing as a community standard and as a reason for malpractice liability if omitted. However, less than half of physicians agreed that radical prostatectomy reduces prostate cancer–specific mortality or disagreed that treatment harms might outweigh treatment benefits, two statements directly addressing patient outcomes from screening.
These results suggest that beliefs unrelated to direct clinical patient outcomes were significant determinants of primary care physician PSA screening practices in 1993 and 1998. This statement is supported by findings that only approximately half of even high testing physicians felt that prostate cancer treatment benefits outweighed treatment harms or that aggressive early treatment in the form of surgery reduced disease-specific mortality. If belief in improved patient outcomes were driving PSA testing practices, more physicians reporting high levels of testing should have supported the efficacy of prostate cancer treatment. Standard of care belief, malpractice concerns, interpretation of PSA test operating characteristics, and markers of community versus academic practice appear to be more highly associated and may be more influential drivers of PSA screening use.
These conclusions are substantiated by our subgroup analysis of physicians who did not endorse the efficacy of early prostate cancer aggressive treatment. This analysis indicated that high testing physicians were much more likely than low to moderate testers to believe that PSA tests represent standard practice and have acceptable test characteristics for disease screening. The strong inverse relationship between overall PSA testing and patient demand–driven testing negates the possibility that patient demand explains the high PSA testing practices of the majority of physicians, indicating instead that patient demand plays an important role for low PSA testing physicians.
The high PSA testing rates observed in this study are consistent with evidence from other studies demonstrating self-reported PSA testing rates from 55% to 87%.12–18 These data corroborate earlier reports that PSA testing is associated with the belief that it represents a standard of practice or common practice,16,17 that PSA tests are ordered based on fear of lawsuits,17 and that PSA tests are acceptable to physicians.17 Our results contrast with other studies that have shown that a primary16 or important reason12,17 for PSA testing was belief that aggressive treatment would reduce mortality or that screening would reduce mortality. The 1998 survey showed a trend favoring an association between physician gender and testing, as have 4 of 5 recent U.S. surveys.12,13,15,16,18 Other studies have shown that physician age12,17 and number of years since graduation from medical school18 were both associated with greater PSA testing, associations also noted in our study. Community versus academic practice location, a major determinant of test ordering in our study, has also been noted by studies in Ohio17 and Oklahoma.16
To illuminate the observed differences in PSA testing, several potential explanations merit consideration. In our study population, high testing physicians were generally located in the community-based practices and low testing physicians were located in the academic practices. The community physicians were older, more likely to be male, and had practiced a greater number of years than had academic physicians. Given the multi-collinearity of these explanatory variables and the small sample size, it was not possible to determine their independent effect on testing by multivariable analysis. However, other surveys have demonstrated similar age and gender relationships with PSA testing, so these may represent unique associations, particularly since gender-associated variation in primary care physician screening practices has been well described.19,20
A potential explanation for the strong association between academic versus community physician practice and PSA testing is that the academic-based physicians may be more likely to incorporate the principles of disease screening and evidence-based medicine into their calculus of medical decision making. This tendency might result in a higher threshold among academic physicians to adopt new technology or a delay before embracing new screening methods such as PSA testing. This idea is consistent with the concept of “‘evangelists’ (advocates of screening)” and “‘snails’ (advocates of the scientific method)” described by Sackett and Holland21 and applied to prostate cancer by Collins and Barry.22 In our 1993 survey, the “evangelists” are the high testing, almost exclusively community physicians and the “snails” are a mix of low testing, mostly academic physicians. Our 1998 data suggest no reduction in enthusiasm for screening on the part of the “evangelists.” Consistent with a lag phase, half the initial “snails” reported a change to a moderate or high testing frequency and acknowledged patterns of attitudes more favorably disposed to PSA testing in 1998. However, the pattern of belief changes we observed, along with the lack of interval-published data in support of PSA screening, suggest that the “converted snail” behavior was not a result of evidence-based review but more likely due to a growing influence of standard of care beliefs and other non–patient outcome–related factors.
The apparent paradox of low belief in treatment efficacy yet high screening practice on the part of some physicians may be due to grouping physician uncertainty and disagreement about early treatment efficacy together for analysis purposes. Disaggregating these response categories revealed that in the presence of uncertainty about efficacy, community and academic physicians adopted opposite testing strategies: 84% of uncertain community physicians adopted high testing strategies, whereas 88% of uncertain academic physicians adopted low or moderate strategies. The findings were similar in both surveys. These relatively high testing rates among community physicians in the face of uncertainty about treatment efficacy support our conclusion that other factors are more influential drivers of PSA test ordering, particularly among community physicians. It also confirms a higher threshold among academic physicians for adoption of a new screening modality.
The fundamental requirements of an effective screening program are well established.23,24 Prostate cancer is an important health problem that can be diagnosed in a presymptomatic stage, two key requirements for effective screening. However, another key principle is that screening programs are only effective when treatment at the presymptomatic stage improves patient mortality or other important outcomes, a point still unsettled in the case of prostate cancer. Given the other potential influences on physician behavior we have described, our findings could be explained if some physicians elect to screen patients widely before these fundamental principles of screening are satisfied. Primary care physicians might choose to limit their role to identification of presymptomatic disease, deferring decisions about treatment appropriateness to specialists. If so, these physicians, who are essentially delinking the utilities of disease identification and treatment, are adopting a more circumscribed posture that diminishes their role in disease screening. This approach might be expected to produce less–well-informed but more frequently tested patients, a hypothesis that may warrant future investigation.
Alternately, physicians may choose to test a high proportion of patients because suggestions to incorporate principles of informed consent and shared decision making may be too difficult to implement in the hectic pace of primary care. Studies of informed decision making show that patients are substantially less likely to elect PSA testing if provided with detailed information about the pros and cons of testing.25–28 Unfortunately, all the methods used in these studies are cumbersome to apply in busy primary care practices. To the extent that these time pressures may weigh most heavily on high testing community physicians, academic physicians may have a greater opportunity to use shared decision making techniques that would be expected to produce lower PSA testing rates. This explanation is also consistent with our finding that low testing physicians are more likely to test as the result of patient demand.
There are several potential limitations to this study. The physician sample was fairly small and was drawn from a large metropolitan area serving an insured, nonindigent patient base with a high managed care penetration. Thus, generalizability to other settings may be limited. Furthermore, community physicians in this study generally received reimbursement on a mixed discounted fee-for-service and capitated basis while academic physicians received salaried compensation from the medical center. While physician estimates of managed care penetration were not associated with PSA testing, confounding may be introduced by failure to account for more subtle differences in payment mechanism. Finally, physicians were asked to self-report their estimated PSA testing rate. The extent to which this reflects their actual PSA ordering behavior is not known.
Our results suggest that beliefs unrelated to direct clinical patient outcomes are important determinants of physician PSA testing practices and that such testing has continued to proliferate and diffuse in recent years. To foster informed patient decision making about prostate cancer screening, physicians need clear and credible data with respect to disease and quality-of-life outcomes in this area. Clinicians need objective guidelines based on such data instead of the current contradictory guidelines based on differing perspectives. Physicians and patients would also benefit from financial and organizational mechanisms that foster a collaborative decision-making process. Without such evidence, tools, and support, physicians will continue to make screening decisions on the basis of perceived imperatives rather than actual patient preferences and objective data.
The authors would like to thank Andrew M.D. Wolf, MD for his careful review and thoughtful suggestions.