|2002, Volume 5, Number 2, Pages 115-118|
|Table of contents Previous Article Next [PDF]
|Probability of prostate cancer at various levels of per cent free prostate specific antigen in Japanese men with total PSA of 4.1-10.0 ng/ml|
|S Egawa1, K Suyama1, R Takashima1, S Kuwao2 and S Baba1|
1Department of Urology, Kitasato University School of Medicine, Sagamihara, Kanagawa, Japan
2Department of Pathology, Kitasato University School of Medicine, Sagamihara, Kanagawa, Japan
Correspondence to: S Egawa, Department of Urology, Kitasato University School of Medicine, 1-15-1 Kitasato, Sagamihara, Kanagawa 228-8555, Japan. E-mail: firstname.lastname@example.org
Estimates for the likelihood of prostate cancer at different levels of per cent free prostate specific antigen (PSA) were derived from experience with consecutive Japanese male patients with intermediate total PSA values who underwent ultrasound-guided biopsies and/or transurethral resection of the prostate. Receiver operating characteristic (ROC) curve analysis showed that in patients with a total PSA of 4.1-10.0 ng/ml, per cent free PSA identified those with prostate cancer better than did total PSA; per cent free PSA also proved superior in the subgroup whose glands appeared benign on palpation. The probabilities of prostate cancer at per cent free PSA values of £10.0, >10-15, >15-20, >20-26 and >26% were 58.3, 40.8, 25.3, 14.3 and 7.6%, respectively, when analyzed without regard to findings on palpation. In patients with palpably benign glands, the corresponding values were 55.3, 35.4, 19.6, 9.7, and 4.6%, respectively. These probabilities are lower than those reported in Western countries, probably reflecting both different patterns of practice and racial differences. Race-specific assessment is recommended before applying a clinical test.
Prostate Cancer and Prostatic Diseases (2002) 5, 115-118. doi:10.1038/sj.pcan.4500562
prostate cancer; total prostate specific antigen; per cent free prostate specific antigen; ultrasound-guided biopsy; racial difference
The diagnostic value of prostate specific antigen (PSA) for differentiating prostate cancer from benign prostatic conditions is limited by its lack of specificity. Numerous investigators have reported the usefulness of per cent free PSA for enhancing cancer detection in men, particularly those with total PSA between 4.1 and 10.0 ng/ml.1,2,3,4 Most of these authors used statistical methods to demonstrate an increase in specificity and a decrease in the number of unnecessary biopsies. Probability estimates, however, are always easier to understand and are more familiar to patients than other ways of presenting information regarding their situations.5 Several Western investigators reported probability estimates for prostate cancer based on parameters including per cent free PSA values and assessed the corresponding need for prostate biopsy.6,7,8 The results of studies carried out in Western countries, however, may not directly translate into Japanese practice owing both to the different patterns of practice and to racial differences in the incidence of prostate cancer.
In view of these considerations, we studied the probability of prostate cancer in Japanese males with intermediate PSA values as affected by values of per cent free PSA.
Materials and methods
Patients and PSA assay
From January 1999 through to July 2000, 372 consecutive men underwent transrectal ultrasound-guided biopsy at Kitasato University Hospital. None of these patients had previously been diagnosed as having prostate cancer or undergone endocrine manipulation. All patients underwent biopsy if their digital rectal examination (DRE) results appeared questionable and/or their serum total PSA was greater than 2.0 ng/ml by Dainapack AxSYM PSA assay (Dinabot, Tokyo).9 Details of these patients have been reported elsewhere; this study represents a separate analysis on the same database.10 Most of these patients had some urinary tract symptoms, thus representing the setting of urologic practice rather than screening population.
After obtaining written informed consent, blood samples were collected by venipuncture immediately before ultrasound-guided biopsy. Serum samples were stored at 4°C for 2-3 h, then were centrifuged at 1500 G for 30 min. Samples were frozen in small portions, stored at -70°C for no more than 3 months and thawed immediately before analysis. All patients underwent ultrasound-guided systematic needle biopsy at six or more separate sites selected so as to maximize sampling of the peripheral zone. Fifteen patients underwent transurethral resection of the prostate gland (TURP) following clinical diagnosis of benign prostatic hyperplasia based on negative biopsy results.
Total and free PSA were measured in serum using the Dainapack AxSYM PSA assay.9 Per cent free PSA was calculated in each sample using the formula: (free PSA concentration/total PSA concentration)´100. Further analysis was limited to patients with total PSA levels of 4.1-10.0 ng/ml (157 patients); a separate analysis was conducted for patients with palpably benign glands (132 patients).
The Kruskal-Wallis test was performed to assess differences among patients within specific per cent free PSA ranges with regard to age, total PSA, and volumes of the transition zone and the gland. The Spearman rank correlation test was used to examine the relationship between per cent free PSA and transition zone or gland volumes. The significance of total PSA and per cent free PSA for predicting prostate cancer was assessed using receiver-operating characteristic (ROC) curve analysis and employing a statistics graphic data measurement software program (STATA computing resource, Los Angeles, CA, USA). P-values <0.05 were considered significant.
A logistic regression analysis was applied to the data from patients with a total PSA level of 4.1-10.0 ng/ml and to the subgroup with palpably benign glands. The binary dependent variable was the biopsy result (cancer or absence of cancer) and the initial independent variables were total PSA, age and ranges of per cent free PSA. A stepwise model selection procedure was carried out, with P<0.05 defined as significant. The probability of prostate cancer within per cent free PSA ranges was calculated using the final logistic model. Sensitivity and specificity within per cent free PSA ranges were also calculated. In addition, the odds ratios were calculated from the initial logistic model for all possible independent predictors of prostate cancer. Pearson 2 goodness-of-fit test was used to test the internal validity of the model.
Patient characteristics and results of biopsy
The median ages (±standard error) of these patients with a total PSA level of 4.1-10.0 ng/ml and of the subgroup with palpably benign glands were 68±0.6 and 68±0.6 y, respectively. The median sizes of the transition zone and of the prostate gland were 18.1±1.2 and 35.9±1.6 cm3, and 18.8±1.3 and 37.4±1.8 cm3, respectively. The per cent free PSA decreases as the transition zone or gland volume decreases (P<0.0001, Table 1).
There were 37 prostate cancers in patients with this total PSA range, of whom 25 had palpably benign glands. Classification of these 37 cancers (numbers in parentheses refer to the 25 cancers in palpably benign glands) identified 1 (1) as stage T1b, 23 (23) as stage T1c, 3 (0) as stage T2a, 6 (1) as stage T2b, 3 (0) as stage T2c and 1 (0) as stage T3a. No distant metastasis was found in 33 (21) of the cases (M0), while no information on metastasis was available in 4 (4) (Mx). One patient was diagnosed as having stage T1b disease as a result of TURP following negative biopsy. This patient was included in the study as a cancer patient. Tumor grade distribution was as follows: Gleason score 2-4 in 9 (5), score 5 or 6 in 18 (15), score 7 in 9 (4) and score 8 in 1 (1).
ROC curve analysis comparing performance of total PSA and per cent free PSA
ROC curve analysis showed that the performance of per cent free PSA in identifying patients with prostate cancer was significantly better than that of total PSA both in the total group of patients with intermediate PSA values and in the subgroup with palpably benign glands (P<0.05).10 Areas under the ROC curves (AUCs) for per cent free PSA and total PSA in the whole group and the subgroup were 0.7653 and 0.6204, and 0.8076 and 0.6479, respectively.
Prostate cancer probability by ranges of per cent free PSA
Age and per cent free PSA were the only significant predictors in the final logistic model after the stepwise model selection procedure (P<0.02 and <0.0001, respectively). Results of the Pearson 2 goodness-of-fit test indicated that this model fits well, thus confirming its internal validity. Calculation of the odds ratios from the initial model indicated per cent free PSA as the strongest predictor of prostate cancer (odds ratio 2.34). The calculated odds ratio for age was 1.1.
The distribution of cancer probability (with 95% confidence limits) for various ranges of per cent free PSA is presented in Table 2. The probability of cancer increases as the per cent free PSA decreases. The sensitivity was lower and the specificity was higher as the per cent free PSA decreased. The probabilities of prostate cancer in the per cent free PSA ranges of £10.0, >10-15, >15-20, >20-26 and >26% were 58.3, 40.8, 25.3, 14.3 and 7.6%, respectively, in the entire group of patients with total PSA of 4.1-10.0 ng/ml. The corresponding probabilities were 55.3, 35.4, 19.6, 9.7 and 4.6%, respectively, in the subgroup with palpably benign glands.
Use of per cent free PSA has been shown to improve specificity in the detection of prostate cancer, particularly in the intermediate range of total PSA (4.1-10.0 ng/ml).1,2,3,4 This finding was confirmed in this study.
Several investigators in Western countries have reported the usefulness of predicted probabilities based on parameters including per cent free PSA values.6,7,8 They recommend using these values to predict the likelihood of cancer in specific patients, rather than using single cut-off values derived from population studies. One advantage is that estimating prostate cancer probabilities on the basis of per cent free PSA may eliminate the debate over setting a predefined 'cut-off' for all per cent free PSA determinations.6,7,8 For example, a per cent free PSA cut-off off 22% would have detected 90% of our subjects with prostate cancer and would have eliminated 47.9% of unnecessary biopsies in the total group of patients with intermediate total PSA values (4.1-10.0 ng/ml). Patients may easily become frustrated as they struggle with the interpretation of such information. Probability estimates based on per cent free PSA ranges can make it easier for the patient to grasp his own risk of having cancer and facilitate discussion of the need for further examination.
A study can provide erroneous estimates of the cancer probability associated with a test if the underlying percentage of cancer patients in the study population does not match that in the population of interest.2,11 The incidence of prostate cancer in this study was lower by at least 10% than the reported incidences in Western series,1,2,4,6,12 23.6% (37/157) for the total patient population and 18.9% (25/132) for those with palpably benign glands. The results of a study carried out in Western countries may thus not directly translate into Japanese practice owing both to the different patterns of practice and to racial differences in the incidence of prostate cancer. We have previously shown that per cent free PSA failed to distinguish patients with prostate cancer from those with benign histology when total PSA was in the lower range (2.1-4.0 ng/ml) in this same Japanese patient cohort.10 This is most likely due to the much lower incidence of prostate cancer in this particular PSA range (approximately 7%).2,10,11 Per cent free PSA may not be applicable to Japanese patients with total PSA below 4.0 ng/ml.
Vessella et al reported that the probability of prostate cancer based on per cent free PSA ranges in a prospective cohort of 297 men at least 50 y old with a total PSA level of 4.1-10.0 ng/ml and nonsuspicious DRE findings.6 These patients were enrolled at 10 clinical sites. The probability estimates for prostate cancer in the per cent free PSA ranges of £10.0, >10-15, >15-20, >20-26 and >26% were 69, 51, 32, 18 and 9%, respectively. The corresponding values in our patients with palpably benign glands were lower: 55.3, 35.4, 19.6, 9.7 and 4.6%, respectively. Indeed, their estimated values were higher than those for our entire group of Japanese patients, without regard to DRE findings (58.3, 40.8, 25.3, 14.3 and 7.6%). This difference in probability may reflect current practice patterns in the two countries.
One of the problems with the per cent free PSA test is that it requires two separate assays, thus increasing labor and cost of diagnosis. The use of a single assay with comparable efficacy may be a solution to this problem. Brawer13 and Maeda et al14 evaluated the usefulness of complexed PSA determined by the Bayer Immuno 1 complexed PSA assay and by a Japanese complexed PSA assay, respectively. Complexed PSA was found to perform better than total or per cent free PSA for the early detection of prostate cancer in men with PSA in the range 4.1-10.0 ng/ml. Interestingly, they later reported in a multicenter trial that complexed PSA alone significantly enhanced diagnostic performance compared to total PSA in the PSA range of 6.01-10.0 ng/ml but not in the PSA range £6.0.12 In the total PSA range of 4.01-10.0 ng/ml, there was no statistically significant difference in specificities between total PSA and complexed PSA. Per cent free PSA significantly outperformed complexed PSA in the 4.01-6.0 total PSA range. Thus, to our knowledge, the usefulness of complexed PSA in this PSA range, either alone or as relative values, is controversial and still remains to be established in a prospective study.12,15,16 The use of a neural network may possibly minimize the uncertainty in this area.17
Interpretation of the results of our study still calls for caution, owing to the small number of patients and the possible bias introduced by employing a single institution's experience. Further investigation based on a prospective multi-institutional study with a larger number of patients is warranted.
We thank WA Thomasson, PhD, for expert editorial assistance. This work was supported in part by a Grant from the Ministry of Health and Welfare of Japan (7-42) and Foundation for Promotion of Cancer Research in Japan.
1 Catalona WJ et al. Use of the percentage of free prostate-specific antigen to enhance differentiation of prostate cancer from benign prostatic disease. A prospective multicenter clinical trial. JAMA 1998; 279: 1542-1547. MEDLINE
2 Van Cangh PJ et al. Free to total prostate-specific antigen (PSA) ratio is superior to total-PSA in differentiating benign prostate hypertrophy from prostate cancer. Prostate 1996; 7: ((Suppl)) 30-34.
3 Egawa S et al. The ratio of free to total serum prostate specific antigen and its use in differential diagnosis of prostate carcinoma in Japan. Cancer 1997; 79: 90-98. Article MEDLINE
4 Van Cangh PJ et al. Free to total prostate-specific antigen (PSA) ratio improves the discrimination between prostate cancer and benign prostatic hyperplasia (BPH) in the diagnostic gray zone of 1.8 to 10 ng/ml total PSA. Urology 1996; 48: 67-70. MEDLINE
5 Egawa S et al. A study on pretreatment nomogram to predict pathological stage and biochemical recurrence after radical prostatectomy for clinically resectable prostate cancer in Japanese men. Jpn J Clin Oncol 2001; 31: 74-81. MEDLINE
6 Vessella RL et al. Probability of prostate cancer detection based on results of a multicenter study using the AxSYM free PSA and total PSA assays. Urology 2000; 55: 909-914. Article MEDLINE
7 Carlson GD, Calvanese CB, Partin AW. An algorithm combining age, total prostate-specific antigen (PSA), and percent free PSA to predict prostate cancer: results on 4298 cases. Urology 1998; 52: 455-461. Article MEDLINE
8 Veltri RW, Miller MC. Free/total PSA ratio improves differentiation of benign and malignant disease of the prostate: critical analysis of two different test populations. Urology 1999; 53: 736-745. Article MEDLINE
9 Dowell B et al. Development of the Abbott AxSYM free PSA assay: performance characteristics and preliminary clinical evaluation. Anticancer Res 1997; 17: 3037-3038. MEDLINE
10 Egawa S et al. Prospective evaluation of prostate cancer detection by prostate specific antigen-related cancer detection by prostate specific antigen-related parameters: comparison in serum and plasma samples. J Urol (in press).
11 Egawa S et al. Results of transrectal ultrasound-guided biopsies and clinical significance of Japanese prostate cancer. Jpn J Clin Oncol 1998; 28: 666-672. MEDLINE
12 Okihara K et al. Comparative analysis of complexed PSA, free PSA and their ratio in detecting prostate cancer. J Urol 2001; 165A: 99.
13 Brawer MK et al. Measurement of complexed PSA improves specificity for early detection of prostate cancer. Urology 1998; 52: 372-378. Article MEDLINE
14 Maeda H et al. Complexed prostate-specific antigen and its volume indexes in the detection of prostate cancer. Urology 1999; 54: 225-228. Article MEDLINE
15 Kikuchi E et al. Prostate specific antigen adjusted for transition zone volume. The most powerful method for detecting prostate carcinoma. Cancer 2000; 89: 842-849. Article MEDLINE
16 Stamey TA, Yemoto CE. Examination of the 3 molecular forms of serum prostate specific antigen for distinguishing negative from positive biopsy: relationship to transition zone volume. J Urol 2000; 163: 119-126. MEDLINE
17 Babaian RJ et al. Performance of a neural network in detecting prostate cancer in the prostate-specific antigen reflex range of 2.5 to 4.0 ng/ml. Urology 2000; 56: 1000-1006. Article MEDLINE
Table 1 Clinical parameters related to per cent free PSA ranges
Table 2 Probability of prostate cancer based on per cent free PSA ranges
|Received 10 September 2001; revised 26 October 2001; accepted 14 November 2001|
|2002, Volume 5, Number 2, Pages 115-118|
|Table of contents Previous Article Next [PDF]