Prostate-specific antigen kinetics contributes to decision making for biopsy referral: the predictive implication for PSA retest in patients with elevated PSA levels

Abstract

Background

It is common to repeat prostate-specific antigen (PSA) measurements for men with PSA elevation before prostate biopsy. In this scenario, they may have considerable psychological distress in fear of the presence of cancer until retests. We assessed possible clinical factors causing transient PSA rise and explored the parameters predictive of subsequent PSA change.

Methods

As interfering conditions, the history of ejaculation, bicycling, and any types of infections were assessed using the questionnaire. The pattern of PSA change was compared in association with the various clinical factors. Predictive significance of PSA kinetics such as coefficient of variation (CV) and PSA velocity (PSAV) for PSA values at retest was evaluated.

Results

The rate of reversion to the normal range was 38.3% at retest. The rate of 12.8% of men showed a large increase by ≥20%, whereas 38.2% of men showed a large decline by ≥20% from the baseline. Men with younger age (≤60 years), small prostate (<20 cc), and prior history of ejaculation or infections showed significantly larger PSA decrease than their counterparts. Those with large CV or PSAV before the baseline more frequently showed PSA decrease below the age-specific cutoff or decline by ≥10% from the baseline at retest. These parameters associated with PSA kinetics had independent predictive values for relevant PSA change at retest.

Conclusions

Ejaculation and any types of infections should be avoided before PSA tests. Men with large PSA fluctuation before the baseline are likely to show a significant PSA decrease at retest. This predictive information may help both physicians to determine whether to proceed to an immediate biopsy and patients to reduce their psychological burden.

1. Introduction

Prostate-specific antigen (PSA) is central in the diagnosis and monitoring of prostate cancer. However, the main limitation to the accuracy of PSA is its inherent variability associated with many factors that are considered to cause transient elevations unrelated to cancer.¹ That is, PSA changes over a short interval are attributed to the variation in the assay, reflecting the sum of analytical and biological variations. Analytical variation mainly results from laboratory processing and assay performance. Biological variation results from individual factors including diurnal and circadian variation, physical and sexual activity, urinary tract infection, digital rectal examination (DRE), prostate biopsy, and administration of alpha-reductase inhibitors.^{2, 3, 4, 5} The latest meta-analysis of 12 studies yielded a mean biological PSA variation of 20% (2.1–22.9%) over a period of days to months.⁶ This leads to a weakness of PSA screening in terms of the considerable incidence of reversion of abnormal to normal results. As such, several clinical practice guidelines recommend repeated PSA measurements to confirm its elevation over the diagnostic threshold before proceeding to a prostate biopsy.^7,8 On the other hand, an abnormal PSA level revealed by PSA screening, suggestive of the presence of prostate cancer, can generate considerable anxiety in asymptomatic men.^9,10 In this context, establishment of the predictive parameters that discriminate transient PSA rise from persistent rise, more suggestive of malign origin may help men with elevated PSA levels relieve their psychological distress.

In this study, we investigated the role of individual factors including physical and sexual activity that influence PSA results and variation over time in an attempt to explore the clinical parameters predictive of subsequent PSA kinetics for biopsy decisions in men referred for elevated PSA levels.

2. Materials and Methods

2.1. Patients and biopsy scheme

We prospectively collected 609 men aged 50 years or older with elevated PSA levels by PSA screening from April 2018 to August 2021 at our hospital. Those under the use of finasteride or dutasteride and men who had urologic events (urinary retention, rectal examination, urine catheterization, cystoscopy, etc.) before PSA measurement were excluded from the study. Patients with confirmed PSA elevation at retest underwent magnetic resonance imaging (MRI) of the prostate. Prostate biopsies were performed not only in those with images suggestive of cancer and PSA level above the normal level but also in those with persistent PSA elevation on repeated measurements or abnormal DRE findings despite negative MRI findings. Usually, 16–20 cores were taken depending on MRI findings and prostate volume by transperitoneal approach in a prostate biopsy. The history of ejaculation or bicycling within 2 days and any types of infections including genitourinary tract infections determined by urine analysis, respiratory tract infection (common cold, flu, bronchitis, and pneumonia), tonsillitis, otitis, and herpes zoster infection within a week before PSA measurements were assessed using the questionnaire (Table S1).

2.2. Assay for PSA

PSA measurements were carried out by chemiluminescence immunoassay (LUMIPULSE Presto PSA, FUJIREBIO, Tokyo, Japan). The manufacture's treated sera were measured as quality control materials and monitored on daily basis to confirm the assay stability. The coefficient of variation (CV) was calculated by dividing the standard deviation of the repeated values by the mean values. Patients' sera were collected and assayed on the same day using the same lot of reagents. We adopted age-specific reference ranges as cutoff PSA levels: 3.0 ng/mL for 64 years and younger，3.5 ng/mL for 65–69 years, and 4.0 ng/mL for 70 years and older.¹¹ PSA velocity (PSAV: ng/mL/year) was calculated by dividing the difference between the two values by the measurement interval.

2.3. Evaluation of the clinical parameters and statistical analysis

DRE, prostate size measurement, and questionnaire collection were done on the referral day. The PSA levels were assessed not only on the reference date (PSA₀ as baseline) but also serially before (PSA_-1, PSA_-2, PSA_-3) and after (PSA₁, PSA₂, PSA₃) the reference date. PSA retest (PSA₁) was done 3 to 4 months apart from the reference date in principle. % PSA changes from the baseline were calculated as (PSA₁−PSA₀) × 100/PSA₀ (%) and classified as the PSA change pattern: stable PSA₁ within 10% change, PSA₁ increased by 10–20%, PSA₁ increased by ≥20%, PSA₁ decreased by 10–20%, or PSA₁ decreased by ≥20%. PSA changes were compared in association with age, prostate volume, histology, recent history of ejaculation, infection, and bicycling. Prostate volume was calculated using the ellipsoid formula (0.52 × length × width × height) in which each dimension was measured on abdominal ultrasonography.

Data were expressed as median (interquartile range), as the Kolmogorov-Smirnov test showed a non-normal distribution of all the data in this study. The intragroup difference was analyzed by the Wilcoxon rank sum test. Each pairwise group difference was analyzed using the Mann-Whitney U test or Kruskal-Wallis test. Differences in distribution for categorical variables were tested by Chi-square test or Fisher exact test. The Spearman rank correlation coefficient was used to assess correlations between various variables. Univariate and multivariate logistic regression analyses were carried out using clinical variables. Receiver operating characteristic (ROC) curve analysis was used to select the threshold value based on the sensitivity and specificity of each value of a given variable. A P value < 0.05 was considered to indicate statistical significance. All statistical analyses were performed using the free R statistical software (version 3.2.2, https://cran.r-project.org/).

The institutional review board reviewed and approved the study protocol (approval number: 18-005).

3. Results

3.1. PSA measurements and changes at retest

The mean analytical coefficients of variation (CV) were 1.63% at 0.63 ng/mL and 1.10% at 14.41 ng/mL, respectively. The median patients' age was 60 years (range, 54–66 years). As indicated in Table 1, most patients underwent serial PSA measurements before and after the reference date of the baseline value. The baseline PSA values increased significantly with patients’ age (r = 0.302, P < 0.001). PSA retest was performed in 517 men after a median of 134 days (range, 117–182 days) from the reference date. The baseline PSA levels (4.26 [3.56–5.46]) decreased significantly at retest (3.70 [2.63–4.93]) (P < 0.001). MRI was carried out in 174 men with elevated PSA levels, of whom 112 men had suspicious findings of cancer and 95 men agreed to undergo prostate biopsy. Finally, 65 men (68.4%) were diagnosed with cancer.

Table 1.

PSA and its related parameters

PSA-1 (ng/ml)	2.75 (1.90–3.43)	455∗
PSA0 (ng/ml)	4.26 (3.56–5.46)	609∗
PSA1 (ng/ml)	3.70 (2.63–4.93)	517∗
Interval between PSA-1 and PSA0 (d)	365 (344–425)
Interval between PSA0 and PSA1 (d)	134 (117–182)
CV between PSA-1 and PSA0 (%)	30.2 (15.0–61.1)
CV between PSA0 and PSA1 (%)	13.9 (6.6–38.9)
PSAV from PSA-1 to PSA0 (ng/ml/y)	1.19 (0.59–2.59)
PSAV from PSA0 to PSA1 (ng/ml/y)	−1.04 (−5.03–0.65)

CV, coefficient of variation; PSA, prostate-specific antigen; PSAV, PSA velocity.

^∗Number of cases.

The details of PSA change at retest are displayed in Table 2. The rate of reversion to the normal range was 38.3% at retest. The rate of 12.8% of men showed a large increase by ≥20%, whereas 38.2% of men showed a large decline by ≥20% from the baseline, 20.2% of which still remained above the age-specific cutoff levels.

Table 2.

Patients’ distribution according to PSA change at retest

PSA change		Relative to the cutoff level
		High	Low
Increased by ≥20%	66	66	0
Increased by 10–20%	48	48	0
Stable within 10% change	139	122	17
Decreased by 10–20%	77	45	32
Decreased by ≥20%	187	38	149
Total	517	319	198

PSA, prostate-specific antigen.

3.2. PSA change at retest in association with the possible clinical factors affecting PSA levels

PSA changes (= PSA₁–PSA₀) were compared between the groups dichotomized by the possible factors affecting PSA levels, including age (by median value), prostate volume (by 20 cc indicative of prostate hyperplasia), histology, recent history of ejaculation, infections, and bicycling. Men with younger age, small prostate, and the history of ejaculation or infections showed significantly larger PSA decrease than their counterparts, whereas histology and prior bicycling did not have a significant effect on PSA change (Table 3, Table 4). The effect of infection on PSA levels was compared between genitourinary tract origin and the others. Significant PSA decreases were seen in both patients with genitourinary tract infection (7.87 [6.91, 10.25] to 2.05 [1.40, 2.90], n = 7) and those with infections from the other origins (4.67 [3.36, 7.50] to 3.10 [2.82, 4.26], n = 24) at retest. However, the magnitude of PSA decrease was greater in the former (−5.28 [−7.67, −5.13]) than that in the latter (−1.52 [−4.22, −0.55]) (p = 0.023).

Table 3.

Comparison of PSA change according to the possible factors affecting PSA levels

Factors		n	PSA change	P
Age (y)	<60	239	−0.68 (−2.18 to 0.08)	<0.001
	≥60	277	−0.22 (−1.30 to 0.55)
Prostate volume (cc)	<20	186	−0.68 (−2.22 to 0.12)	0.017
	≥20	149	−0.42 (−1.31 to 0.29)
Histology	Benign	29	0.50 (−0.32 to 0.70)	0.234
	Cancer	51	0.60 (0.18 to 1.72)
Ejaculation	No	468	−0.24 (−1.66 to 0.33)	0.027
	Yes	49	−1.08 (−2.35 to 0.20)
Infection	No	485	−0.42 (−1.64 to 0.35)	<0.001
	Yes	31	−2.16 (−5.89 to 0.61)
Exercise	No	497	−0.43 (−1.71∼0.33)	0.303
	Yes	20	−0.75 (−2.23 to 0.13)

PSA, prostate-specific antigen.

Table 4.

Univariate and multivariate logistic regression analysis predicting PSA decline by 10% or more at retest

Factors	Univariate analysis			Multivariate analysis
	OR	95% CI	P	OR	95% CI	P
Age	2.090	1.470–2.980	<0.001	1.980	1.230–3.190	0.005
Prostate volume	1.270	0.821–1.950	0.285		–
Ejaculation	1.330	0.764–2.330	0.310		–
Infection	0.232	0.094–0.576	0.002	0.973	0.234–4.040	0.970
Exercise	0.508	0.202–1.280	0.151		–
CV PSA-1/0	0.123	0.078–0.194	<0.001	0.298	0.169–0.524	<0.001
PSAV-1/0	0.121	0.077–0.190	<0.001	0.249	0.141–0.439	<0.001

CI, confidence interval; CV PSA_-1/0, coefficient of variation between PSA_-1 and PSA₀; OR, odds ratio; PSA, prostate-specific antigen; PSAV_-1/0, PSA velocity from PSA_-1 to PSA₀.

3.3. Parameters predictive of PSA change at retest and its application to biopsy outcomes

To explore the predictive parameters for PSA change at retest, we focused on the PSA kinetics before the baseline such as coefficient of variation between PSA_-1 and PSA₀ (CV PSA_-1/0) and PSAV from PSA_-1 to PSA₀ (PSAV_-1/0). These variables were obtained from 518 men without extrinsic factors affecting PSA values such as ejaculation and infections as parameters reflective of biological PSA variation. Then, patients were dichotomized by the threshold values of CV PSA_-1/0 (0.116) or PSAV_-1/0 (1.043) defined by ROC analysis that discriminates between individuals who experienced PSA₁ decrease below the age-specific cutoff or PSA₁ decline by ≥10% and individuals who did not (Fig. 1). Those with smaller CV PSA_-1/0 or PSAV_-1/0 more frequently showed higher PSA values above the age-specific cutoff or showed stable/increased PSA₁, whereas their counterparts were more likely to show PSA₁ decrease below the age-specific cutoff or decline by ≥10% from the baseline (P < 0.001, Fig. 2). Needless to say, PSA decreases were significantly greater in those with larger CV PSA_-1/0 or PSAV_-1/0 compared with their counterparts (P < 0.001, Fig. 3).

Fig. 1.

ROC curves of sensitivity versus specificity of CV PSA and PSAV to determine the threshold for PSA decrease below the age-specific cutoff (A) and decline by ≥10% from the baseline (B). CV, coefficient of variation; PSA, prostate-specific antigen; PSAV, PSA velocity; ROC, receiver operating characteristic.

Fig. 2.

Comparison of case distribution according to the PSA kinetics before the baseline and PSA values at retest.

Patients were dichotomized by the threshold values of CV PSA or PSAV before the baseline defined by ROC analysis that discriminate between individuals whose PSA levels at retest decreased below the age-specific cutoff or decline by ≥10% and individuals who did not. Then, the proportion of individuals was compared according to the PSA values at retest whether they were higher or lower than the age-specific cutoff, or whether they decreased by > 10% from the baseline using the Fisher exact test.

(A) Proportion of patients stratified by CV PSA before the baseline (threshold: 0.328) and PSA levels at retest relative to the age-specific PSA cutoff.

(B) Proportion of patients stratified by the PSAV before the baseline (threshold: 1.023) and PSA levels at retest relative to the age-specific PSA cutoff.

(C) Proportion of patients stratified by CV PSA before the baseline (threshold: 0.280) and the PSA change at retest.

(D) Proportion of patients stratified by PSAV before the baseline (threshold: 1.273) and the PSA change at retest. CV, coefficient of variation; PSA, prostate-specific antigen; PSAV, PSA velocity; ROC, receiver operating characteristic.

Fig. 3.

Comparison of PSA change according to the PSA kinetics before the baseline.

(A) PSA change according to CV PSA before the baseline

(B) PSA change according to PSAV before the baseline. CV, coefficient of variation; PSA, prostate-specific antigen; PSAV, PSA velocity.

To test the predictive power of CV PSA_-1/0 and PSAV_-1/0 for PSA change at retest, multivariate analysis was performed together with the aforementioned clinical factors, showing that the proposed PSA kinetics before the baseline were the independent predictive parameters for PSA decrease by ≥10% at retest (Table 4).

Thus, we applied such PSA kinetics to the prediction of biopsy outcomes. As only men with PSA above the age-specific cutoff levels at retest were applied to a prostate biopsy, the incidence of cancer was similar regardless of whether they showed PSA decline by ≥10% or not. Likewise, ROC analysis determined the threshold values of the postbaseline PSA kinetics such as CV PSA_0/1 or PSAV_0/1 that discriminates cancer and benign histology. Proportion of patients dichotomized by these thresholds did not differ with the biopsy outcomes (Table 5). Also, the prebaseline PSA kinetics was not associated with the biopsy outcomes, either (data not shown).

Table 5.

Comparison of the case distribution according to PSA kinetics and biopsy outcomes

	Noncancer	Cancer
Increase/stable	22	36
Decrease >10%	5	9
∗P		1
Small CV PSA0/1∗∗	17	21
Large CV PSA0/1	10	24
∗P		0.226
Small PSAV0/1∗∗	16	21
Large PSAV0/1	11	24
∗P		0.338

CV, coefficient of variation; PSA, prostate-specific antigen; PSAV, PSA velocity.

^∗P value by Fisher exact test.

^∗∗Patients were dichotomized by the threshold values for CV PSA (0.116) or PSAV (1.043).

4. Discussion

The recent rationale of prostate cancer screening has been focused on reducing unnecessary biopsy along with overdiagnosis and overtreatment based on the evidence that treating men with low-risk prostate cancer led to a deterioration of the quality of life without a risk of increased mortality.¹² In this scenario, several clinical practice guidelines recommend repeating PSA measurement to confirm its elevation over the cutoff level before performing a prostate biopsy.^7,8 Such strategy may help exclude a transient PSA rise caused by intraindividual PSA variation considered as a main factor affecting the accuracy of PSA result for biopsy referral.⁶ Indeed, it was demonstrated that a fourth of men with referral PSA levels between 4 and 10 ng/mL had decreased to the normal range at retest, resulting in increased specificity for biopsy referral up to 80% and decreased risk of unnecessary biopsy up to 60%.¹³ On the contrary, one study reported that 71% of men with initially elevated PSA had a reduction, with values <4.0 ng/mL in 38%, although 43% of men with prostate cancer showed a PSA decrease below their baseline levels, suggesting that short-term PSA decrease may not be an excuse to avoid a prostate biopsy.¹⁴ The other study demonstrated that the rate of reversion to normal PSA range at retest was negligible regardless of the later biopsy result (0.9% in cancer and 3.7% in non-cancer patients) and thus PSA retest was not always justified before proceeding to a prostate biopsy.¹⁵

Some activities and events in daily life including bicycling, sexual activity, and infections have been considered to contribute to PSA elevation. Among several reports describing the effect of ejaculation on PSA levels, six studies on elderly men eligible for screening were more relevant to discussing this issue, four of which showed an increase^{16, 17, 18, 19} and two showed no significant effect.^20,21 Tchetgen et al reported that PSA levels significantly increased in 87% of 64 men after ejaculation and returned to normal by 48 hours in 91% of them.¹⁶ Thus, they recommended abstinence from ejaculation for at least 48 hours before PSA measurement, which was the basis of our definition as an interfering condition against PSA tests. Mechanical stimulation of the prostate is a generally accepted interfering factor for PSA measurement²² and long-distance bicycling is a typical activity to study the combined effect of mechanical prostate stimulation and intense endurance exercise.²³ The results from the literature concerning the association between bicycling and PSA levels are inconsistent.^{24, 25, 26, 27, 28} Thus, Jiandani et al conducted a meta-analysis from 8 studies on this issue and demonstrated no significant increase in PSA associated with bicycling.²⁹ However, majority of the studies which did not show a PSA change after bicycling included only younger men under the typical age for PSA screening, of whom the baseline PSA level might be too low to increase beyond interassay CV, rendering it statistically insignificant.^{23, 24, 25, 26} Indeed, bicycling caused PSA increase in the studies restricted to subjects ≥50 years old.^27,28 Moreover, it is not generally known that infections other than urogenital origin such as respiratory tract can also affect PSA levels, although its exact mechanism is unclear.^30,31 As the prostate is considered an immunocompetent organ with a complex intraglandular immune system, defined as the prostate-associated lymphoid tissue and accompanied by physiological inflammation, it may respond to infections beyond the prostate itself, resulting in PSA elevation.³² In fact, we observed a significant PSA decrease at retest in patients with a recent history of nonurogenital infections before the initial measurement. Despite this knowledge, there is no specific warning in daily practice that calls for abstinence from ejaculation and strenuous bicycling before PSA tests or even avoidance of PSA tests during and for a while after any infectious disease. Actually, 101 men (16.6%) had either of those events before the baseline PSA test.

On the other hand, an abnormal PSA level revealed by PSA screening, suggestive of the presence of prostate cancer, can generate considerable anxiety in asymptomatic men.^9,10 Moreover, many physicians are indecisive as to whether to perform a prompt biopsy when they encounter variations in PSA levels during repeated measurements, which may accelerate patients’ psychological distress on screening. As such, if there is a predictor that can differentiate whether the present PSA elevation will be transient or persistent, it would be of great use for both physicians to determine whether to proceed to an immediate biopsy or stay on repetitive PSA measurements and patients to reduce their psychological burden. We used the increased or decreased PSA cutoff value of 10% and 20% based on the report that 20% decrease in PSA values is associated with a reduced risk of cancer and particularly of high-grade cancer.³³ Thus, a certain degree of decrease in PSA values should be meaningful for the patients even if their PSA levels are still above the normal range at retest.

In the present study, a significant fraction of patients showed PSA reversion to normal range (in 38.3%) or large decline by ≥ 20% from baseline (in 36.2%) at retest. Among the possible interfering events, ejaculation and any types of infections but not bicycling were likely to cause PSA increase. If patients with a recent history of these events had been recognized beforehand and excluded from PSA testing, they may have been spared the anxiety brought by inaccurately caused PSA elevation. Furthermore, after excluding the aforementioned interfering conditions, we attempted to seek the predictors of PSA change at retest, which may be attributed to randomly occurring intraindividual variation. We focused on the PSA kinetics just before the baseline such as CV and PSAV. Patients with smaller CV PSA or PSAV more frequently remained above the age-specific cutoff or showed stable/increased PSA at retest, whereas their counterparts were more likely to show decrease below the age-specific cutoff or decline by ≥10% from the baseline. Although CV PSA and PSAV showed similar predictive significance for subsequent PSA values, these parameters have different meaning; the former indicates variation relative to the mean and is useful to compare the scatter of two differently scaled quantities (i.e., PSA changes of 0.8 to 1.2 ng/mL and 8.0 to 12.0 ng/mL yield the same CV), while the latter encompasses a time component. To put it simply, men with large PSA fluctuation before the baseline may have a more chance for significant PSA decrease at retest. Many physicians may intuitively understand this idea in daily practice, but to our knowledge, this is the first study showing it as objective data.

We attempted to apply this formula to predict the biopsy outcomes. However, postbaseline as well as prebaseline PSA kinetics could not efficiently discriminate cancer and benign histology because only men remaining above the age-specific cutoff levels at retest were indicated for a prostate biopsy in whom PSA variability cannot be large under such biopsy protocol. On the other hand, Park et al reported the predictive feasibility of the pattern of PSA fluctuation for biopsy outcomes in the repeat biopsy setting. That is, patients with PSA fluctuations and PSAV ≥1.0 ng/ml/yr had about 3-fold higher risk of prostate cancer compared with those with continuous increase in PSA or those having PSA fluctuations with PSAV <1.0 ng/ml/yr.³⁴

We should note a couple of limitations of the present study. First, the history of ejaculation, infection, and bicycling before PSA measurements was based on the questionnaire obtained retrospectively and thus may not always ensure sufficient accuracy. Second, we did not exclusively choose urogenital infections and bicycling but incorporated various types of infections and strenuous bicycling into the possible factors affecting PSA levels despite the lack of firm evidence. Although bicycling had little effect on PSA levels, significant PSA decreases were seen at retest in men having infections even from nonurogenital origins. This novel finding should be confirmed in a larger number of patients. Third, our biopsy protocol recommended a biopsy to patients with persistent PSA elevation and thus the exact association between PSA kinetics and biopsy outcomes is not fully clarified, especially when PSA decreased below the age-specific cutoff. Lastly, the proposed cutoff of CV PSA and PSAV that predict PSA kinetics at retest should be confirmed in a larger cohort of patients or validated by other studies. Such efforts may more reliably specify who should retest their PSA levels and who should proceed to prostate biopsies without a retest.

In conclusion, more than a third of men with elevated PSA levels experienced reversion to normal range at retest. Any types of infections as well as ejaculation before PSA tests are likely to cause PSA rise, and thus specific warning to avoid these interfering events before PSA measurements must be advocated in daily practice. Intraindividual variation is another major source of transient PSA increase, indicative of implicit possibility of cancer, which can generate psychological distress. Men with large PSA fluctuation indicated by CV or PSAV before the baseline are likely to have a more chance for significant PSA decrease at retest. This predictive information may help both physicians to determine whether to proceed to an immediate biopsy or stay on repetitive PSA measurements and patients to reduce their psychological burden in fear of the presence of cancer.

Conflicts of interest

All authors have no conflict of interest to declare.

Appendix A. Supplementary data

The following is the Supplementary data to this article: