ABSTRACT: An association between age and prostate-specific antigen (PSA) has been documented: As men age, their serum PSA value increases. Currently, a single demarcation between normal and elevated PSA values, 4.0 ng/mL, is used as an indication for biopsy among men of all ages. The use of age-specific reference ranges might address several shortcomings of the PSA test. First, age-specific reference ranges could improve the sensitivity of PSA by detecting curable, organ-confined tumors in younger men when the threshold of 4.0 ng/mL is lowered. Second, age-specific reference ranges could improve the specificity of the PSA test by raising the PSA threshold for normal among older men. This would modulate PSA interpretation by taking into account the increasing prevalence of both benign prostatic growth and incidental, non-life-threatening cancers among successively older cohorts of men. Furthermore, many unnecessary (false-positive) biopsies would be avoided. However, the association between PSA values and age is not entirely clear, and whether age-specific reference ranges represent the best interpretive index for PSA remains problematic. [ONCOLOGY 11(4):475-485, 1997]
Prostate-specific antigen (PSA) is the best "tumor marker" yet discovered.[1,2] This enzyme is produced by the columnar epithelial cells of the prostate and periurethral glands. As a serine protease in the kallikrein gene family, PSA presumably has some biologic function in the prostate or in its secretions.
Semenogelin, a major protein in seminal fluid, is cleaved by PSA, and this cleavage is apparently an important part of the liquefaction of semen. Prostate-specific antigen also cleaves one of the six binding proteins of the insulin-like growth factor, IGF BP-3. In vitro studies have shown that although IGF BP-3 inhibits the activity of IGF-II, its cleavage by PSA reverses this inhibition and frees IGF to stimulate proliferation.[4,5] Unfortunately, far less is known about other biologic substrates of PSA.
The basement membrane of the acini, basal cells lining the acini, and stromal cells act as barriers to prevent the escape of PSA into the bloodstream. Therefore, serum levels of PSA are normally maintained below 4 ng/mL, which corresponds to about 10-6 of the levels of PSA in seminal fluid.
Normal prostate epithelial cells and benign hyperplastic tissue actually produce more PSA protein than does malignant tissue, and PSA messenger RNA (mRNA) is also expressed at higher levels in benign tissue than in malignant prostate tissue.[7,8] Therefore, PSA is not a true tumor marker. Abnormalities in the architecture of the prostate gland resulting from either trauma or disease cause increased "leakage" of the enzyme into the bloodstream via capillaries and lymphatics.
Prostate-specific antigen is specific to the prostate but not to prostate cancer. Prostate cancer can cause a breakdown of the barriers that prevent escape of PSA into the extracellular fluids, resulting in elevated PSA levels in the blood. Prostate-specific antigen levels increase approximately in proportion to the volume of prostate cancer. However, elevated serum levels of PSA do not always result from prostate cancer. Benign conditions, such as bacterial prostatitis, urinary retention, and benign prostatic hyperplasia (BPH), may also cause elevations in serum PSA levels. Although PSA concentrations increase with the volume of BPH tissue, the average increase is small (0.3 ng/mL/g of tissue) when compared with the increase associated with cancer (nearly 3.5 ng/mL/g of malignant tissue).
Prior to the era of PSA testing, the digital rectal examination was the gold standard for the detection of prostate cancer. However, interoperator variance with the digital rectal examination has been documented; its sensitivity is low; and it is more likely to detect locally advanced cancer than organ-confined disease.
The PSA test also is not a flawless method of cancer detection. The major shortcoming of the PSA test is its less-than-perfect sensitivity and specificity rates, despite the fact that these rates are among the highest of cancer screening tests currently in use. A relatively high number of false-positive results is common: PSA levels above normal but resultant negative biopsies. In most screening studies, the positive predictive value of a PSA level 4.0 ng/mL or more is approximately 33%.[13-16]
Combined use of PSA testing and digital rectal examination is universally recommended because of the significant increase in positive predictive value. Approximately 25% to 33% of patients with prostate cancer have serum PSA concentrations in the normal range at the time of diagnosis, made on the basis of a digital rectal examina-tion or a PSA threshold less than 4.0 ng/mL.[15-17] Only 1 patient in 10 with a suspicious digital rectal examination will prove to have prostate cancer on biopsy if the PSA value is less than 4.0 ng/mL.[17,18] When the PSA value is 10.0 ng/mL or higher, at least 50% of patients will have prostate cancer. Between these two groups is the diagnostic "gray zone" of the PSA value, 4.0 to 10.0 ng/mL, for which a positive biopsy of cancer is problematic. In this range, nearly 25% of men will have prostate cancer, and 75% will have benign prostatic enlargement. In fact, approximately 25% of all men with a histopathologic diagnosis of BPH have serum PSA levels higher than commonly used cut-off value of 4.0 ng/mL.
Various analytic methods have been proposed to improve the sensitivity and specificity of PSA testing. Higher standards are necessary but not sufficient proof that PSA testing will reduce mortality from prostate cancer. These indexes include calculations of PSA density, PSA density adjusted for volume of the transition zone, analysis of PSA velocity, molecular forms of PSA,[23,24] and the application of age-specific PSA reference ranges.
The calculation of PSA density--PSA values in relation to the volume of the prostate gland--was initially proposed by Benson and colleagues as a method to differentiate BPH and cancer. Calculation of PSA density necessarily involves a transrectal ultra- sonographic examination to derive an estimation of volume, and the operator dependency of this examination can influence subsequent calculations of prostatic volume.
Studies of PSA density have reported mixed findings: Some investigators[26,27] have found that PSA density calculation improves the sensitivity and specificity of PSA testing, whereas others have reported no diagnostic improvement with this calculation. Refinements in the interpretation of PSA density calculation have been attempted, such as consideration of the volume of the transition zone; volume-referenced PSA values; and an unpublished study of age-specific PSA density. However, no consistent diagnostic improvement can be claimed for PSA density and its derivatives. A significant correlation between age and volume of the prostate gland has been documented, and this phenomenon supports the consideration of age-specific reference ranges.[25,30]
A landmark analysis, based on the Baltimore Longitudinal Study of Aging, demonstrated a significant difference in the age-adjusted rate of change in PSA levels among groups of men who had prostate cancer, BPH, and no prostate disease. Carter and associates suggested that a 0.75-ng/mL/yr increase in PSA is predictive of a diagnosis of cancer. However, intraindividual variation of PSA test results, the lack of standardization of PSA assays, and the recommendation that a minimum of three annual tests over a 2-year period be used to calculate PSA velocity, have hampered the establishment of a "normal" PSA change (slope, or velocity) over time. An association between age-specific reference ranges and PSA velocity seems to be intuitively apparent, but long-term prospective studies of such an association have not yet been reported.
Different Molecular Forms of PSA
Considerable enthusiasm has been generated by the possibility of differentiating benign and malignant disease through the assay of the proportion of different molecular forms of PSA: (1) PSA bound to alpha-1-antichymotrypsin, (2) PSA bound to alpha-2-macro-globulin, and (3) unbound (free) PSA. However, establishing a "correct" ratio of these different PSA types that will distinguish prostate cancer from benign growth will likely be debated, much as PSA and its indexes are debated currently. Although the amounts of these different molecular forms and of total PSA apparently vary according to age, the ratios do not, and may ultimately obviate any age-related consideration in diagnostic evaluations.
Studies Establishing Age-Specific Ranges
Early studies of PSA testing for the detection of prostate cancer found consistent, strong associations between PSA values and glandular volume, PSA values and age, and age and glandular volume.[14,30,33] Brawer and colleagues found that the mean PSA value increased with age (P = less than .0001), but the correlation coefficient (r)--the degree of correlation between age and PSA values--was only .15. (It is important to note that when the correlation coefficient is squared [r2], resulting in a coefficient of determination, that percent explains the linear relationship between the two variables. In the case of the findings of Brawer's group, this would mean that 2.3% of the variation in PSA levels can be explained by its linear relationship to age).
Babaian and colleagues also found a significant association between PSA levels and age and a significant relationship between prostatic volume (determined by transrectal ultrasonography) and increasing PSA levels and age (for all, P less than .00005). The increasing prevalence of both benign prostate disease and prostate cancer as men age had long been reported, and PSA seemed to be related to this phenomenon. This study recommended biopsy when the PSA value is between 4.0 to 10.0 ng/mL and the prostatic volume is 25 ccor less.
Collins and associates reported on a clinic population referred for assessment of BPH. The study's objective was to investigate the relationships between prostatic volume and PSA levels and between age and prostatic volume. Correlations were established, but the study also found an independent association between PSA levels and age. Linear regression analysis showed an independent association between PSA levels and age when adjustments were made for prostate volume. Age and prostate volume influence PSA levels independently. A modest correlation of PSA levels and age was reported (r = .37; P less than .0001).
The seminal study of age-specific reference range by Oesterling and associates reported on a community-based population (N = 471) that had no evidence of prostate cancer. Prostate-specific antigen levels correlated with age (r = .43 [r² = 18.5]; P less than .0001); a higher correlation of PSA with PSA density was found (r = .55; P less than .0001); and PSA density was weakly correlated with age (r = .25; P = less than .001). With the median PSA value plus two standard deviations, age-specific reference ranges were established for 10-year age groups. These age-specific reference ranges have become the standard in the literature (Table 1). Based on regression analysis of the cross-sectional data, this study estimated that the serum PSA concentration increases approximately 3.2% per year. For a healthy, 60-year-old man, that would mean an increase of .04 ng/mL over the next year.
Oesterling has been at the forefront of advocating the use of age-specific reference ranges to make PSA a more discriminating tumor marker for detecting clinically significant cancers in older men (increasing specificity by raising the threshold for normal PSA levels) and potentially curable cancers in younger men (increasing sensitivity by lowering the threshold for normal PSA levels).[34,35] Other studies[36,37] have investigated this association between PSA levels and age, and similar age-specific reference ranges have been proposed (Table 1).
The age-specific reference ranges of Oesterling's group are relatively conservative, given the generally lower ranges in each age group. This may reflect less PSA variability among the relatively homogeneous male population of Olmstead County, Minnesota. Their age-specific reference ranges may reflect a standard against which "abnormally" high PSA levels can be evaluated. The incidence of prostate cancer may also be lower in Olmstead County because of the number of research studies conducted among its population. As increasing percentages of men undergo PSA testing, and as more cancers are detected, regression to the mean PSA values of Olmstead County may occur throughout the country.