Complexed prostate-specific antigen for the diagnosis of biochemical recurrence after radical prostatectomy


J. Kellogg Parsons, UCSD Division of Urology, 200 West Arbor Drive, San Diego, CA 92103, USA. e-mail:



To determine the validity of using complexed prostate-specific antigen (cPSA) levels for diagnosing biochemical recurrence after radical prostatectomy (RP).


With linear regression modelling, we determined threshold cPSA levels for biochemical recurrence in patients after RP for clinically localized prostate cancer. We calculated sensitivity, specificity, predictive values, and likelihood ratio tests of each threshold for diagnosing biochemical recurrence using total PSA (tPSA) as the reference standard.


In the regression models, tPSA and cPSA were highly correlated (r = 0.99). For the diagnosis of biochemical recurrence, tPSA thresholds of 0.20 and 0.40 ng/mL corresponded to cPSA thresholds of 0.12 ng/mL (95% confidence interval 0.08–0.17) and 0.29 (0.22–0.28)  ng/mL, respectively. For the detection of biochemical recurrence, a cPSA threshold of 0.12 ng/mL had a sensitivity of 96%, specificity of 88%, positive predictive value of 89%, negative predictive value of 88%, positive likelihood ratio of 8, and negative likelihood ratio of 0.05; the respective values for a cPSA threshold of 0.29 ng/mL were 96%, 96%, 96%, 96%, 24 and 0.04.


cPSA has high validity for the diagnosis of biochemical recurrence after RP. Pending external validation, cPSA might be useful for biochemical surveillance after RP.


radical prostatectomy


(complexed) (total) PSA.


Compared to total PSA (tPSA), serum complexed PSA (cPSA) has better specificity for detecting prostate cancer at all clinically relevant sensitivities [1–15]. The use of cPSA as a single serum marker for prostate cancer could potentially improve the specificity for cancer detection while simultaneously minimizing the redundancy, confusion and costs associated with the performance and interpretation of multiple PSA isoform assays [16]. However, there are only limited data on the validity of using cPSA for clinical functions other than prostate cancer screening.

One of these clinical functions is the detection of biochemical recurrence after radical prostatectomy (RP). With an estimated incidence in the USA of several thousand cases per year and a steadily increasing prevalence [17], biochemical recurrence after RP is an important, frequently encountered issue in clinical urology. Each year, tens of thousands of men undergo surveillance with tPSA assay after RP.

Preliminary data suggest that cPSA might be useful for detecting biochemical recurrence after definitive treatment for localized prostate cancer [18]. Therefore, we investigated the validity of the serum cPSA assay for the diagnosis of biochemical recurrence after retropubic RP.


Institutional Review Board approval was obtained for this study. Single serum specimens were studied in 150 men who had retropubic RP for clinically localized prostate cancer at a single tertiary-care centre, and who had detectable serum concentrations of tPSA (0.03–1.08 ng/mL) after surgery. Specimens had been stored for up to 2 months at − 20 °C. In addition, 144 longitudinal specimens from 15 men previously diagnosed with biochemical recurrence for a tPSA level of ≥ 0.2 ng/mL and followed for a mean (sd, range) of 4.2 (1.4, 2–6) years after surgery were analysed. Specimens had been stored at −70 °C for 1–15 years before analysis.

PSA and cPSA were analysed using the ADVIA Centaur (Bayer HealthCare, Tarrytown, NY, USA); the analytical sensitivity of the two assays, as defined by the manufacturer, are 0.01 and 0.03 ng/mL for tPSA and cPSA, respectively. Between-run assay imprecision was determined to be 3.9–4.5% (0.25–0.40 ng/mL) and 5.5–6.4% (0.11–0.20 ng/mL) for tPSA and cPSA, respectively.

The primary outcomes were the correlation of cPSA with tPSA; the determination of cPSA thresholds equivalent to tPSA thresholds of 0.2 and 0.4 ng/mL; and sensitivity, specificity, predictive values, and likelihood ratios of cPSA thresholds for detecting biochemical recurrence. The secondary outcome was the correlation of cPSA with tPSA in monitoring patients diagnosed with biochemical recurrence.

In 50 patients, linear regression modelling was used to calculate the correlation between cPSA with tPSA levels, and to predict cPSA values corresponding to two commonly used tPSA thresholds for biochemical recurrence, i.e. 0.20 and 0.40 ng/mL [17]. In an additional 100 patients, the sensitivity, specificity, predictive values and likelihood ratios for each cPSA threshold were then determined with 2 × 2 table analysis (25 cases with tPSA greater than the threshold and 25 controls with tPSA less than the threshold) using each of the tPSA thresholds as the reference standard.

In 15 patients with biochemical recurrence, cPSA and tPSA concentrations in 144 serial serum samples (tPSA 0.02–47.4 ng/mL) were compared using linear regression modelling with robust variance estimates to account for multiple measures over time in the same individuals.


Among 50 patients with a detectable tPSA level of <1.0 ng/mL the mean (sd, range) tPSA level was 0.58 (0.16, 0.34–0.98) ng/mL and the mean cPSA was 0.36 (0.12, 0.17–0.67) ng/mL. tPSA and cPSA correlated highly (r = 0.99, P < 0.001; Fig. 1). In the regression model, tPSA values of 0.2 and 0.4 ng/mL predicted cPSA values (95% CI) of 0.12 (0.08–0.17) and 0.29 (0.22–0.28) ng/mL, respectively.

Figure 1.

Regression analysis of cPSA on tPSA in 50 patients after retropubic RP at one academic centre (r = 0.99, P < 0.001).

With a predicted cPSA of 0.12 ng/mL the mean tPSA and cPSA for the 25 controls was 0.15 (0.03, 0.08–0.19) and 0.09 (0.03, 0.03–0.15) ng/mL, respectively. The mean tPSA and cPSA for the 25 cases was 0.31 (0.06, 0.21–0.45) and 0.19 (0.06, 0.07–0.35) ng/mL, respectively. The sensitivity, specificity and predictive values were generally high (Tables 1 and 2).

Table 1.  A 2 × 2 table comparing cPSA to tPSA for diagnosing biochemical recurrence after retropubic RP in 25 cases and 25 controls from one academic centre at two thresholds of tPSA and cPSA
cPSA, ng/mL tPSA, ng/mL
Recurrence YesNoYesNo
≥0.12Yes24 3  
No 122  
≥0.29Yes  24 1
No   124
Table 2.  Measures of the validity of cPSA for detecting biochemical recurrence after RP at one academic centre
MeasuretPSA, ng/mL
cPSA, ng/mL 0.12 0.29
cPSA range, ng/mL 0.03–0.35 0.05–0.68
Sensitivity, %9696
Specificity, %8896
Positive predictive value, %8896
Negative predictive value, %9296
Likelihood ratio:
 Positive 8.024.0
 Negative 0.05 0.04

With a predicted cPSA of 0.29 ng/mL, the mean tPSA and cPSA for the 25 controls was 0.24 (0.06, 0.13–0.33) and 0.15 (0.05, 0.07–0.27) ng/mL, respectively. The mean tPSA and cPSA for the 25 cases was 0.64 (0.15, 0.43–0.92) and 0.42 (0.14, 0.18–0.68) ng/mL, respectively. The sensitivity of cPSA at a threshold of 0.29 ng/mL was equivalent to that for 0.12 ng/mL, but the specificity and predictive values were higher (Tables 1 and 2).

In 15 patients diagnosed with biochemical recurrence longitudinal estimates of cPSA and tPSA correlated highly (r > 0.99, P < 0.001; Fig. 2); this analysis included serial PSA levels in one patient with biochemical recurrence treated with androgen ablation 54 months after surgery.

Figure 2.

Longitudinal values of cPSA and tPSA in a patient with biochemical recurrence after RP who was treated with androgen ablation therapy 54 months after surgery.


In this cohort of prostate cancer patients treated with RP, cPSA levels correlated highly with tPSA levels and had high validity for the diagnosis of biochemical recurrence. These data suggest that cPSA might serve as an alternative to tPSA for biochemical surveillance after RP, and raises the possibility that cPSA might also be applicable to surveillance after primary radiotherapy, cryosurgery and androgen ablation.

In one previous analysis of patients with early- and late-stage prostate cancer, which included those treated with RP, there was a high concordance of serial cPSA levels with clinical status in 97% of men [18]. By contrast, in the present study we determined the validity of cPSA as a serum marker for biochemical recurrence by using tPSA as the reference standard.

These data also show, for the first time, the applicability of cPSA to a routine clinical function other than prostate cancer screening and diagnosis. The specificity of cPSA for prostate cancer detection is comparable to the combination of tPSA and free PSA [12]. Replacing several serum tests with a single PSA isoform would probably limit confusion and redundancy while optimizing prostate cancer detection and reducing unnecessary prostate biopsies. Indeed, the potential now exists for cPSA to fulfil all of the roles currently performed by tPSA, including screening and diagnosis, prognostication in staging nomograms [19], and longitudinal monitoring of patients with prostate cancer.

Although the present validity of cPSA was generally higher at a threshold of 0.29 ng/mL (corresponding to a tPSA level of 0.40 ng/mL) than of 0.12 ng/mL (tPSA 0.20 ng/mL), the use of separate PSA ranges for analysing separate thresholds precluded direct validity comparisons between thresholds. We used different PSA ranges to restrict the analyses to clinically relevant values around each threshold (Table 2). The use of a single range would have incorporated PSA values much greater than the threshold, which would probably have increased the sensitivity for detecting biochemical recurrence. Instead, we chose a more conservative approach to maximize the potential external validity of the results.

We used several different measures to estimate the clinical validity of cPSA. The positive predictive value (the proportion of patients testing positive who actually have the disease) is a particularly useful measure, but depends in part on the prevalence of the disease in the study population; the higher the prevalence, the greater the positive predictive value [20]. The prevalence of biochemical recurrence in this selected group was 50%. As biochemical failure typically occurs in about a third of patients after RP it is possible that the positive predictive value would be less in an unselected clinical series of patients.

Another caveat on the results is that that the estimated cPSA thresholds for biochemical recurrence (0.12 and 0.29 ng/mL) are preliminary. External validation in other study populations should be done before routine use in clinical practice.

In conclusion, cPSA has high validity for detecting biochemical recurrence after RP. Pending external validation, cPSA might be useful for biochemical surveillance after RP.


Carol Cheli is an employee of the sponsor. Source of funding: Bayer Pharmaceutical.