Empiric antibiotics for an elevated prostate-specific antigen (PSA) level: a randomised, prospective, controlled multi-institutional trial

Authors


Correspondence: Scott Eggener, 5841 South Maryland, Mail Code 6038, Chicago, IL 60637, USA.

e-mail: seggener@surgery.bsd.uchicago.edu

Abstract

Objective

  • To determine the impact of empiric antibiotics on men with an elevated prostate-specific antigen (PSA) level.

Subjects/Patients and Methods

  • Men of any age with a PSA level of >2.5 ng/mL and normal digital rectal examination undergoing their first prostate biopsy were recruited from five medical centres.
  • Patients with previous biopsy, prostate cancer, urinary tract infection (UTI) or prostatitis within the prior year, antibiotic use within 1 month, 5α-reductase inhibitor use, allergy to fluoroquinolones or clinical suspicion of UTI were excluded.
  • Men were randomised to 2 weeks of ciprofloxacin or no antibiotic. A PSA measurement was obtained 21–45 days after randomisation immediately before prostate biopsy.
  • The primary endpoint was the change in PSA level between baseline and immediately before biopsy.

Results

  • Complete data were available for 77 men with a mean (interquartile range) age of 60.6 (53–66) years.
  • In the control group of men not receiving antibiotic (39 men), the mean baseline and pre-biopsy PSA levels were 6.5 and 6.9 ng/mL, respectively (P = 0.8).
  • In men receiving ciprofloxacin (38 men), the mean baseline PSA level was 7.6 ng/mL and after 2 weeks of ciprofloxacin was 8.5 ng/mL (P = 0.7).
  • Compared with controls not receiving antibiotic, use of ciprofloxacin was not associated with a statistically significant change in PSA level (P = 0.33).
  • Prostate cancer was detected in 36 (47%) men, 23 (59%) in the control group and 13 (34%) in the antibiotic group (P = 0.04). Detection rates were not significantly associated with the change in PSA level between baseline and biopsy. The primary limitation of the study is early stoppage due to an interim futility analysis and poor accrual.

Conclusion

  • Despite not meeting the target accrual goal, empiric use of antibiotics for asymptomatic men with an elevated PSA level does not appear to be of clinical benefit.

Introduction

Worldwide, millions of prostate biopsies are performed annually [1, 2], most commonly due to an elevated PSA level. Numerous non-cancer aetiologies can cause an elevated PSA level, including benign prostatic hyperplasia (BPH), inflammation, and infection.

The use of empiric antibiotics for a newly elevated PSA level is common. When presented with the clinical scenario of a man with a newly elevated PSA level, a recent survey of non-urology physicians revealed nearly a third would initiate a trial of empiric antibiotics [3].

Typically, patients receive 1–2 weeks of either fluoroquinolone or sulpha-based antibiotics to treat a presumptively spurious elevation of PSA level caused by subclinical prostatitis. After a course of antibiotics, the PSA measurement is repeated and if it remains elevated, urological referral and biopsy are recommended. If it significantly decreases, a biopsy may be avoided. This strategy lacks a strong foundation of evidence and antibiotic use in this setting is associated with drug-related side-effects [4], promotion of microbial resistance [5], and an increased rate of sepsis after prostate biopsy [6, 7].

Previous studies have reported PSA level declines following empiric antibiotics [8-12], but nearly all those studies were limited by the lack of a control group and/or lack of randomisation. Therefore, it is unclear if the observed PSA level changes were due to natural variation, regression toward the mean, or a true effect of the antibiotic [13]. We hypothesised that empiric antibiotics for men with an elevated PSA level without clinical evidence of infection are not beneficial. To address this question, we performed a randomised prospective study evaluating PSA level changes in a cohort of men treated with and without empiric antibiotics before undergoing an initial prostate biopsy.

Subject/Patients and Methods

Institutional Review board approval was obtained at five academic medical centres for accrual of men with a PSA level of >2.5 ng/mL and a normal digital rectal examination (DRE) undergoing their initial TRUS-guided prostate biopsy. Previous prostate biopsy, history of prostate cancer, UTI or prostatitis within the prior year, use of 5α-reductase inhibitors, antibiotic use ≤1 month prior to baseline PSA level, allergy to fluoroquinolones, or clinical suspicion for urinary tract infection (UTI) were exclusion criteria. Men underwent urine analysis or urine culture, if clinically indicated.

Men were randomised to either 2 weeks of ciprofloxacin, 500 mg twice daily, or no antibiotic. Randomisation was centralised at a single study site using the permuted blocks method [14]. TRUS-guided prostate biopsy consisted of a minimum of 12 cores. A short prophylactic institution-specific course of antibiotic was started the day before the biopsy. Prostate biopsy was performed 21–42 days after randomisation (7–28 days after completing ciprofloxacin) and PSA measurement was repeated immediately before prostate biopsy. The primary endpoint was change in PSA level. Secondary endpoints included histological findings, relationship of PSA level change to biopsy results, and incidence of infectious complications.

Continuous variables were compared using a two sample t-test and categorical variables compared using Fisher's exact test at the two-sided α level of 0.05. Given the skewness typical of PSA levels, these data were analysed on the logarithmic scale. Sample size calculations indicated that to have 90% power to detect a 20% decrease in PSA levels for subjects treated with antibiotics compared with no change in the control group, using a one-sided level of significance of 0.05, a sample size of 88 subjects (44 per group) was required. However, due to poor accrual, an unplanned interim analysis was performed after 70 patients had complete data. For the interim evaluation of efficacy, an O'Brien-Fleming boundary with z-value of 2.80 for statistical significance was used. For the futility assessment, a stochastic curtailment procedure was used. Early termination for futility was predetermined if the conditional power for rejecting the null hypothesis given the interim data was <10%. This rule would result in a small power loss of no >2%. Interim analysis revealed a conditional power under the alternative hypothesis of 44%. This finding, combined with poor accrual, led to early termination of the trial.

Results

Complete data were available for 77 men with a mean (interquartile range) age of 60.6 (53–66) years (Table 1). In the control group, not receiving antibiotic (39 men), the mean (sd) baseline PSA level was 6.5 (3.0) ng/mL. The repeat PSA level after a period of observation was 6.9 (7.5) ng/mL. In the group receiving ciprofloxacin, the mean (sd) baseline PSA level was 7.6 (7.5) ng/mL. After 2 weeks of ciprofloxacin the repeat PSA level was 8.5 (9.9) ng/mL. The mean change in PSA level, analysed on a logarithmic scale to reduce skewness, from baseline to biopsy was not significantly different between the two groups (P = 0.33; Table 2).

Table 1. Demographics
VariableTotalObservation groupAntibiotic group
Number of subjects773938
Race, n (%)   
 Caucasian49 (64)23 (59)26 (69)
 African-American22 (28)12 (31)10 (26)
 Other6 (8)4 (10)2 (5)
Mean (interquartile range) age, years60.6 (53.1–66.7)61.7 (55.5–66.7)60.4 (53.1–66.6)
Mean (range) PSA level, ng/mL7.06.5 (2.4–17.4)7.6 (3.1–45.7)
Table 2. Biopsy data
VariableObservation groupAntibiotic groupP
Number of Subjects3938 
Mean (sd) PSA level at randomisation, ng/mL6.5 (3.0)7.6 (7.5)0.8
Mean (sd) PSA level at biopsy, ng/mL6.9 (7.5)8.5 (9.9)0.7
Mean log PSA level change between randomisation and biopsy–0.080.0050.33
Prostate cancer at biopsy, n (%)23 (59)13 (34)0.04

Overall, prostate cancer was detected in 36 (47%) men (Table 2). In the observation group, 23 of 39 (59%) men were diagnosed with prostate cancer, while 13 of 38 (34%) in the antibiotic group were diagnosed with cancer (P = 0.04). Of the patients who had an interval PSA level increase, 19 of 36 (53%) had biopsy detectable prostate cancer compared with 16 of 40 (40%) who had an interval PSA level decrease (P = 0.35). One patient had no change in PSA level between measurements.

In the observation group, cancer detection rates did not differ based on interval PSA level increase [12 of 20 (60%)] or decrease [11 of 19 (58%)] (P = 0.9). In men receiving antibiotics, cancer was detected in seven of 16 (44%) with an interval PSA level increase compared with five of 21 (24%) with an interval PSA level decrease (P = 0.3).

In the observation group, three of 39 (8%) patients had a repeat PSA level of ≤2.5 ng/mL of whom one had cancer and 10 of 39 (26%) had a repeat PSA level of ≤4.0 ng/mL with four diagnosed with cancer. In the antibiotic group, one of 38 (3%) patients had a PSA level of ≤2.5 ng/mL with no resultant cancer and four of 38 (10%) had a PSA level of ≤4.0 ng/mL, with one harbouring cancer.

Regarding the degree of PSA level decrease, in the observation group four of 39 (10%) patients had a reduction of >50% with all four patients having negative biopsies. In the antibiotic group, one of 38 (3%) patients had a reduction >50% and was found to have prostate cancer on biopsy.

Sepsis after biopsy occurred in one (1.3%) patient. This patient was in the antibiotic group and the aetiology of his sepsis was a ciprofloxacin-resistant Escherichia Coli.

Discussion

In this prospective, randomised, controlled, multi-institutional clinical trial of asymptomatic men with a newly elevated PSA level, an empiric 2-week course of ciprofloxacin did not lead to decreases in mean PSA levels and repeat PSA levels did not differ compared with the control observation group. Additionally, prostate cancer detection rates did not differ between men with an increasing vs decreasing PSA level, regardless of randomisation group.

PSA is not specific to prostate cancer and can therefore be elevated due to prostate size or transient conditions, e.g. infection and inflammation [15-17]. Administration of antibiotics has been associated with significant PSA level decreases in patients with chronic prostatitis [18-20], BPH with concomitant prostatitis [21], and asymptomatic prostatitis [8-12].

PSA levels can also naturally vary, as 44% of men with a PSA of >4 ng/mL will subsequently have a PSA level of <4 ng/mL during subsequent visits over the following 4 years [13].

In efforts to avoid unnecessary biopsies, physicians often prescribe a course of fluoroquinolone or sulpha-based antibiotics for an elevated PSA level to empirically treat the possibility of subclinical infection or inflammation [3, 21]. Multiple previous studies have examined empiric antibiotics for men with an elevated PSA level but nearly all have lacked a control group, so it is unclear if the observed PSA level changes were secondary to natural variation, regression toward the mean, or a true antibacterial effect [8-13, 19]. Erol et al. [8] prospectively randomised 97 patients with PSA levels of >4 ng/mL to ciprofloxacin and diclofenac vs placebo and reported a significant decrease in PSA level for the treatment group. However, the study population does not reflect contemporary screened patients as the mean PSA level was 25 ng/mL, over half of the patients had an abnormal DRE, and not all patients underwent biopsy.

Based on the limitations of previous studies, we designed our trial to evaluate contemporary screened patients with a newly elevated PSA level and randomised the antibiotic intervention against a control group. We did not observe any differences in PSA level outcomes whether or not patients received antibiotics. The presumptive ideal response of an empiric trial of antibiotics is to have the PSA level return to a ‘normal’ level. Among patients randomised to antibiotics, only 3% had a repeat PSA level of <2.5 ng/mL and 10% had a repeat PSA level of <4 ng/mL. Based on these data, we question the clinical benefit of empiric antibiotic use for asymptomatic men with an elevated PSA level.

While we observed a very small proportion of men ‘normalising’ their PSA levels after antibiotics, previous studies have evaluated whether ‘normalisation’ of PSA level is associated with a lower risk of prostate cancer detection [12, 19]. Stopiglia et al. [19] assessed men who had an interval normalisation of PSA level to <2.5 ng/dL. Three of 10 patients in the placebo group and five of nine in the antibiotic group had a repeat PSA level of <2.5 ng/dL and were subsequently diagnosed with prostate cancer. Therefore, the authors concluded normalisation of PSA level did not indicate the absence of prostate cancer. Similarly, Baltaci et al. [12] described 17 of 100 (17%) men with PSA levels decreasing to <4 ng/mL after antibiotic therapy; five of the 17 (29%) had prostate cancer on biopsy.

The extent of PSA level decrease after ciprofloxacin has also been proposed to be predictive of biopsy results [10]. Serretta et al. [10] diagnosed prostate cancer in two of 13 men with a PSA level reduction of >50% and none of five men with a reduction of >70%. In the present study, four (10%) men in the observation group had a PSA reduction of >50% and all had negative biopsies. In the antibiotic group, one (3%) man had a reduction of >50% and had prostate cancer on biopsy. Based on the present findings, it appears a PSA level decrease of >50% is relatively uncommon and of limited clinical utility.

Men receiving antibiotics had a statistically significant lower rate of prostate cancer (34% vs 59%). The present study was not specifically designed to investigate this outcome. Nevertheless, it remains an interesting observation worthy of future study.

Not only are empiric antibiotics for an elevated PSA level lacking clinical efficacy, they may lead to side-effects and, perhaps, an increased risk of infection after biopsy. Fluoroquinolones have known gastrointestinal (nausea, diarrhoea), CNS (dizziness, headache, seizures), cardiovascular (prolonged QT interval), musculoskeletal (tendon rupture), endocrine (glucose dysregulation), and renal (nephritis, renal insufficiency) side-effects [4]. Outpatient use of fluoroquinolones has been associated with an increased incidence of ciprofloxacin-resistant, community acquired E. coli UTIs [5]. Most pertinent to patients with an elevated PSA level, patients who received 3 weeks of a fluoroquinolone antibiotic before biopsy had a significantly higher incidence of post-biopsy sepsis (5.4% vs 1.7%) and all patients with bacteria identified on cultures harboured a fluoroquinolone-resistant organism [6]. It is therefore paramount to use antibiotics responsibly in patients with an elevated PSA level.

The present trial does have several limitations. First, the trial examined a single antibiotic agent given for an arbitrary duration that we felt reflected common clinical practice. Longer therapy could, perhaps, result in different outcomes in PSA level changes and biopsy findings. Second, the trial was not completed due to an interim analysis suggesting futility and less-than-expected accrual. It is possible that a larger trial would identify a clinical benefit for empiric antibiotics, although we think this is unlikely based on the power calculations. Third, despite being a randomised trial, the PSA levels at baseline were slightly different although not statistically different. This anomalous finding was solely due to a single patient with a PSA level of 45 ng/mL at baseline and 44 ng/mL after his course of antibiotic. Fourth, repeat PSA levels and prostate biopsy were allowed within a 3-week window (21–42 days after randomisation). It is possible that if all patients had had their PSA level and biopsy done at the same time, e.g. 30 days after randomisation, the present study findings might have slightly differed. Lastly, PSA levels can vary based on the assay and we did not standardise the assay.

To our knowledge, we report the first prospective, randomised, controlled, multi-institutional trial of antibiotics for an elevated PSA level in asymptomatic men. Based on these data, empiric use of antibiotics for an elevated PSA level in an asymptomatic patient does not have clinical benefit. For men with a newly elevated PSA level, tangible evidence of an infection warrants appropriate use of antibiotics. Without evidence of an infection, a repeat value after a period of observation is sensible, in order to confirm the newly elevated level.

In conclusion, despite the present study not reaching its target accrual goal, the use of empiric antibiotics for an elevated PSA level in an asymptomatic patient does not appear to be of clinical benefit.

Conflict of Interest

None declared.

Ancillary