INTRODUCTION
Prostate cancer (PCa) is the most commonly diagnosed visceral cancer in men with transrectal ultrasound-guided biopsy (TRUBP) the most frequently used method of tissue diagnosis [1]. Approximately one million biopsies are performed each year in the USA, with an exponential rise observed over the past decade [Reference Loeb2]. TRUBP is generally considered a safe procedure but is invasive with up to 70% of patients experiencing one or more complications [Reference Djavan3]. The majority of these are minor, self-limiting, non-infectious complications (haematuria, haematospermia, perineal pain). The most significant morbidity of biopsy relates to infectious complications, including urinary tract infection, bacteraemia and sepsis [Reference Loeb2, Reference Roberts4, Reference Williamson5]. These complications are thought to occur due to inoculation of the prostate and surrounding tissues with bacterial flora of the rectal mucosa, most commonly Escherichia coli. Fluoroquinolone (FQ)-based antimicrobial prophylaxis is recommended by many authorities including the American Urological Association and the European Association of Urology [6, 7].
Severe infection is the most common reason for both hospitalization and primary-care intervention following TRUBP and is associated with a substantial health and economic burden [Reference Williamson5]. Indeed, post-TRUBP infection accounts for up to 20% of all E. coli bacteraemias in men, who are twice more likely to require intensive care admission compared to infections acquired in the community [Reference Williamson8]. This may be attributed to higher rates of resistant organisms causing post-biopsy infections, as well as a generally older population with medical co-morbidities undergoing biopsy. The incidence of infectious complications following TRUBP is reported to be increasing worldwide, attributed to increasing antibacterial resistance by organisms causing post-biopsy infections [Reference Loeb2, Reference Carignan9, Reference Nam10]. The most clinically significant phenotypes are those of FQ resistance and extended-spectrum β-lactamase (ESBL) production [Reference Williamson8, Reference Steensels11]. By contrast, this trend has not been observed for non-infectious complications, which have remained stable [Reference Loeb2].
To aid clinicians in their selection and monitoring of TRUBP patients, identification of risk factors for infectious complications have been attempted. Suggested risk factors relate to FQ-resistant bacterial carriage as a result of recent antibiotic use, hospitalization, urological infections, or international travel, diabetes mellitus, and a history of FQ-resistant infection [Reference Roberts4, Reference Carignan9, Reference Steensels11, Reference Simsir12]. Repeated biopsies are indicated in men with persistent suspicion of PCa based on serum prostate specific antigen (PSA) levels or digital rectal examination findings, and are becoming an integral part of the management plan for those electing to have active surveillance of early disease [Reference Simsir12]. Subsequent biopsies have not been associated with an increased risk of infectious complications, although this should be re-evaluated in the context of increasing antimicrobial resistance, with prior TRUBP reported to be a risk factor for colonization with resistant E. coli [Reference Liss13, Reference Zowawi14].
Significant variability in biopsy technique has been reported, potentially due to ongoing debate regarding the optimal strategy [Reference Loeb2]. Factors such as transrectal or transperineal sampling, number of cores, sampling sites, and antimicrobial prophylaxis can influence both quality of the pathological sample as well as rates of post-biopsy complications. Transperineal biopsy (TPBP) was routinely used prior to the 1980s, and is still preferred in some centres in Europe and Asia [Reference Takenaka15]. TPBP is at least as efficacious as TRUBP in PCa detection, and may detect anteriorly sited tumours better [Reference Hossack16, Reference Shen17]. It is, however, associated with an increased logistical and financial burden. As TPBP avoids the ‘dirty to clean’ passage of rectal mucosa, it has traditionally been thought to have lower rates of infection than the ‘transfaecal’ alternative. This benefit is less clear in practice; some studies of TPBP find incidence of sepsis <1% even without prophylaxis, while others report equivalent infection rates to the transrectal route [Reference Takenaka15, Reference Hara18, Reference Miller, Perumalla and Heap19]. Notwithstanding these variations, the major morbidity of TPBP is associated with acute urinary retention requiring hospitalization [Reference Dimmen20].
A lack of prospective data on post-biopsy complications, and inter-country variation in biopsy and prophylactic regimens, means that the true incidence of post-biopsy infection is difficult to determine. Direct comparisons between TRUBP and TPBP are few and have mostly focused on PCa detection rates [Reference Takenaka15, Reference Hara18, Reference Miller, Perumalla and Heap19, Reference Nesi21]. Accordingly, this study aimed to (i) systematically review all of the available literature on post-biopsy infections, (ii) determine the overall burden of the major complications of TRUBP and TPBP, (iii) assess the pattern of regional variation in post-biopsy complications.
METHODS
Data sources
A systematic review of the literature was conducted in August 2013 in accordance with the PRISMA Statement and Cochrane Guidelines [Reference Higgins and Green22, Reference Moher23]. The following databases were included: Cochrane Central Register of Controlled Trials (CENTRAL); Ovid Medline; EMBASE, CINAHL, and LILACS. The search strategy included medical subject headings, synonyms and truncated descriptors for the following terms: prostate, neoplasm, biopsy, infection, culture, bacteraemia, sepsis, fever, urinary tract infections, post-operative complications (Supplementary Table S1). Searches were not restricted by time and non-English citations were excluded. Reference lists of articles undergoing full-text review were manually searched. Citations were stored and categorized using Endnote X6 (Thomson Reuters, USA).
Study selection
Two authors (H.B., M.J.R.) independently screened citations in two rounds. First, studies were screened by title and abstract and duplicates were removed manually. Articles were then reviewed in full text. Any discrepancies between reviewers also resulted in full text review of the article. Eligible for inclusion were randomized trials, cohort studies (prospective or retrospective) or case series, which investigated the incidence of post-biopsy complications. Studies were excluded if they did not specifically state the type of biopsy, and what complications were assessed. Articles were graded for study quality as ‘low’, ‘medium’, or ‘high’ risk of bias using the Hoy Risk of Bias tool [Reference Hoy24].
Data extraction
A standardized form was developed in Excel (Microsoft Corporation, USA) to collect pertinent information from selected studies. Extracted data included study characteristics (design, location, time-course), patient characteristics (demographics, prophylaxis, biopsy parameters, follow-up) and complications (types, time course). The major complications of biopsy seen were acute urinary retention, bacteraemia, bacteriuria, hospitalization, and clinically diagnosed sepsis. Only those studies with culture proven bacteraemia or bacteriuria post-operatively were included. ‘Hospitalization’ referred to all hospitalization related to biopsy, not limited to the other outcome measures. ‘Sepsis’ referred to the systemic inflammatory response syndrome in the context of infection, though was not consistently defined across included studies, with articles reporting either a ‘clinical diagnosis’ or recording the specific criteria.
Statistical analysis
Initial review of the data suggested considerable heterogeneity between studies, with location of study appearing to be an important source [Reference Bennett25]. Due to the expected heterogeneity in true prevalence of complications between countries, a two-step process of data pooling was performed.
In step 1, meta-analysis was used to generate a single within-country prevalence estimate. We used the inverse variance heterogeneity (IVhet) model of meta-analysis to pool within-country estimates because it avoids the major problems of overdispersion and increased mean squared error seen with the random-effects model [Reference Doi26, Reference Doi27].
In step 2, risk adjustment was used to aggregate country-specific complication prevalence proportions across countries (within regions and overall). We used the directly standardized effect estimate (DSE) to aggregate country-specific data within regions or overall [Reference Hodges and Clayton28]. The DSE method uses a meta-analytical approach to achieve direct standardization. It removes inverse variance weighting as this is no longer relevant to varying true effects and implements subpopulation weights (aka standardization or risk adjustment) [Reference Doi, Barendregt and Rao29]. We used subpopulation weights based on country-specific prostate cancer incidence under the assumption that the differing burden of new cases determines the contribution of that country's prevalence of complications to the standardized prevalence estimate [1]. This method uses a quasi-likelihood approach to generate a variance for the standardized estimate that does not suffer from overdispersion [Reference Doi, Barendregt and Rao29]. Random-effects models were not used in step 2 because the mathematical form of a random effect has been advocated here for convenience only and it would not be possible to make a second draw from the same mechanism that produced these study estimates in the first place [Reference Hodges and Clayton28]. We therefore find it hard to imagine how the random-effects method can deliver any inference here or how any form of meta-analysis using precision-based weights can deliver inference across truly different country estimates [Reference Hodges and Clayton28, Reference Doi, Barendregt and Rao29].
All analyses were done on the double arcsine square-root-transformed proportions (to stabilize the variances) and these were back-transformed for reporting [Reference Barendregt30]. Meta-analysis was performed using MetaXL 2.0 (http://www.epigear.com). Publication bias was assessed for within-country meta-analyses, where possible using Doi plots given that funnel plots are unreliable when prevalence proportion is the effect size [Reference Onitilo, Doi, Barendregt and Williams31, Reference Hunter32].
An Excel spreadsheet was designed to compute the standardized prevalence in each of the 32 analyses (eight overall, 24 regions). Statistically significant differences were determined from non-overlap of the confidence intervals and this implied P < 0·05 given that all were 95% confidence intervals.
RESULTS
Description of included studies
Overall, 3952 citations were returned by the search strategy. A total of 575 references were selected for full text review, after exclusion of duplicate and irrelevant citations. This produced 165 articles for inclusion in the final analysis, representing a total of 162 577 patients (Supplementary Table S2). The flow diagram of study selection is illustrated in Figure 1.
PRISMA flow diagram of study selection. From the initial 3952 citations, 165 articles were included in the final analysis.
Included studies were published between 1971 and 2013, and were mainly of a prospective design. Studies were mostly considered to have a ‘low risk’ of bias (n = 102), with ‘moderate risk’ (n = 50), and ‘high risk’ (n = 13) studies also included (Supplementary Table S3). ‘High risk’ studies were those that did not adequately consider case-definition or the appropriateness of the prevalence period. All studies examined at least one of the chosen outcome definitions, with the majority conducted in Asia, Europe, and North America. The large majority of studies featured either ‘2 week’ or ‘1 month’ follow-up, with no studies reporting complications of biopsy beyond 1 month. Insufficient data were available for pooling of bacteraemia and bacteriuria rates for TPBP patients. The four studies directly comparing morbidity of TRUBP and TPBP were of insufficient size to provide a meaningful pooled analysis [Reference Takenaka15, Reference Hara18, Reference Miller, Perumalla and Heap19, Reference Nesi21]
Within-country prevalence estimates were derived from meta-analysis. Only the United States TRUBP studies were of sufficient number to assess publication bias (Supplementary Fig. S1). The funnel and Doi plots demonstrate a clear positive prevalence bias and this was likely a result of both small study effects and possibly unpublished lower prevalence studies. It is likely that this extends to other within-country estimates as well and therefore these results are less conservative than they ideally should be.
Standardized prevalence of major complications following prostate biopsy
The results of our analysis are presented in Table 1. There were no statistically significant differences in prevalence between regions or across biopsy types, as the confidence intervals overlapped for the pooled estimates. TPBP estimates were generally higher for urinary retention than TRUBP (values given in parentheses are 95% confidence intervals) [4·2% (0·2–12·9) vs. 0·9% (0–3·6)], which was consistently observed across Asian, European and North American studies. The prevalence of bacteraemia following TRUBP was also higher in Asian countries compared to North America [3·5% (2·1–5·2) vs. 0·7% (0·2–1·6]). The estimates for bacteriuria following biopsy were similar, with overall prevalence of 7·3% (5·3–9·6) in Asian studies vs. 6·4% (4·5–8·6) in North America. Overall hospitalization was similar for TRUBP compared to TPBP [1·1% (0–3·9) vs. 0·9% (0–3·4)]. Hospitalization was generally more prevalent following TRUBP in Asian studies [2·2% (0–7·7) vs. 0·6% (0·1–1·4)] but uncertainty around pooled estimates was high due to the observed heterogeneity. Sepsis rates were higher for TRUBP than TPBP [0·8% (0–3·0) vs. 0·1% (0–0·2)], and this was consistent across continents, but again lacked precision. Complication rates for individual countries are summarized in Supplementary Figure S2, and reported in Supplementary Table S4.
Standardized estimates of prostate biopsy complications. Standardized prevalence and 95% confidence intervals for overall, Asian, European and North American studies
CI, Confidence interval.
Total values are pooled values (directly standardized effect estimates) for all countries examining that outcome measure. Regional data were derived from risk adjustment of country-specific data. Country-specific data were derived through meta-analysis if there was more than one report. No statistically significant differences were found across transrectal and transperineal sites stratified by region.
DISCUSSION
This study reports a systematic literature review and burden-of-disease analysis on the available literature of the major complications following TRUBP or TPBP. This is the first study to address burden of disease using a method of risk adjustment that enables true variations by geographical region to become evident (which meta-analysis does not allow). The latter accounts for the varying at-risk populations across regions, and our results indicate that TRUBP was associated with higher rates of hospitalization and sepsis than TPBP but lower rates of urinary retention. Asian studies generally reported higher rates of bacteraemia, bacteriuria and sepsis after TRUBP compared to other regions. However, this was not statistically significant as studies were quite heterogeneous.
The risks of infection following TRUBP have long been recognized, with early studies reporting a propensity for fever, urinary tract infection, sepsis and occasionally mortality following biopsy [Reference Thompson33]. Despite use of peri-procedural antibiotics, infections after TRUBP continue to cause significant morbidity. As demonstrated in this study, TRUBP is associated with clinically significant rates of bacteraemia [1·2% (0·2–12·6)], bacteriuria [5·8% (0·2–18·6)], sepsis [0·8% (0–3·0%)], and consequently hospital admission [1·1% (0–3·9)]. In 2013, a Canadian study of 75 190 men found a fourfold increased risk of hospitalization following TRUBP from 1996 to 2005 (1·0% to 4·1%, P < 0·001) [Reference Nam10]. This was reflected in another Canadian study that reported the incidence of infection significantly increased from 0·52% in 2002–2009 to 2·15% in 2010–2011 (P < 0·001) [Reference Carignan9]. Similar increases were found in a sample of 17 472 US men biopsied during 1991–2007 [Reference Loeb2]. The large number of TRUBP performed has also led to an array of rarer disseminated infections, including osteomyelitis, meningitis, infective endocarditis, pyelonephritis, Fournier's gangrene, blindness, and even a combination of the above (Supplementary Table S5). Infections are associated with substantial economic burden, with costs of admission, investigations, extended antimicrobial treatment, and outpatient follow-up after biopsy, estimated to be greater than that for MRSA bacteraemia and Clostridium difficile infections [Reference Batura and Rao34].
Infection rates had dropped substantially with the routine use of antibacterial prophylaxis, but are now considered to be increasing [Reference Kapoor35]. Concurrently, increasing rates of FQ resistance have been documented worldwide since 1990, associated with profligate use of antimicrobial drugs [Reference Johnson36]. In particular, FQ resistance is reported to be high in many Asian populations with rates of 40–70% published, reflecting the ready availability and ‘over the counter’ access to these agents [Reference Apisarnthanarak37, Reference Williamson38]. Our analysis demonstrated higher levels of bacteraemia, bacteriuria and sepsis after biopsy in Asian studies compared to those from Europe and North America. Hospitalization was also more prevalent after TRUBP in Asian countries, which may be linked to the increased burden of resistant organisms causing infection. Faecal carriage of FQ-resistant organisms in men undergoing prostate biopsy was reported to be approximately 20%, and resistant bacteria are responsible for at least 50% of post-biopsy infections in North America [Reference Feliciano39, Reference Taylor40]. Rates of post-biopsy infection may therefore be related to changing patterns of antimicrobial resistance. This reflects multiple factors, including local resistance patterns, prior antibiotic usage, hospitalization, and international travel [Reference Carignan9, Reference Nam10].
Biopsy approach is a major factor influencing morbidity. TPBP avoids the transient bacterial seeding from the rectum that is thought to cause infection in TRUBP. This was reflected in the smaller rates of TPBP sepsis seen in our pooled analysis [0·1% (0–0·2)], and lower risk of fever and sepsis reported in comparison with TRUBP and is likely also to be reflected in the lower hospitalization rates seen after TPBP [0·9% (0–3·4)], which were most evident in Asia where infection rates were highest. As demonstrated by the results of our analysis, the major associated morbidity following TPBP was acute urinary retention. Increased experience with the technique may account for the lower rates of urinary retention, hospitalization and sepsis seen after TPBP in Asian and European countries compared to North American reports. TPBP is logistically more involved and time-consuming, requiring admission to hospital and an operating theatre, and so is substantially more expensive. Given the practical advantages of transrectal biopsy, the current consensus is that the transperineal route should be reserved for those at high risk of sepsis, or for patients suspected of having anteriorly sited tumours [Reference Hossack16]. In future, the emergence of more selective biopsy strategies based on multi-parametric MRI stratification is likely to lead to fewer biopsy procedures being performed, which may make transperineal biopsies a more practical option.
The key limitations of this study are that results are subject to the inherent biases of the predominantly observational study designs, with potential for inconsistent selection of study participants. Regional variation in health systems and availability of primary healthcare likely influenced the threshold for hospital admission between countries, which we could not account for in our study. The impact of many patient factors and biopsy variables could not be analysed systematically, so biopsy technique and geographical region were assessed as the main homogenizing factors. The paucity of English-language studies published in many parts of the world would also influence the results. In particular, the heterogeneity within Asia was unable to be explored with the available data and so the systematic review may underestimate the burden of post-biopsy infection in under-represented regions such as South East Asia. The potential also remains for an array of rare disseminated post-biopsy infections not covered here. These complications are listed in Supplementary Table S5, with prophylactic regimens for special subgroups of patients (such as those with infectious endocarditis) covered in these reports.
In conclusion, widespread use of PSA testing and the rise of active surveillance in PCa management have led to an exponential rise in the number of TRUBP performed internationally over the last decade. This systematic review benchmarks the major morbidity of prostate biopsy internationally with higher rates of infections warranting hospitalization with TRUBP compared to TPBP, although the latter was associated with higher rates of acute urinary retention. By pooling all available data, this is also the largest scale comparison between biopsy techniques. It supports the need for further research directly comparing TRUBP and TPBP morbidity in this era of increasing infections. Finally, use of the DSE approach has enabled incorporation of risk adjustment into the burden of disease analysis, which has been noticeably absent from previous studies.
SUPPLEMENTARY MATERIAL
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0950268815002885.
ACKNOWLEDGEMENTS
No financial support was required for this project. M.J.R. is supported by a Doctor in Training Research Scholarship from Avant Mutual Group Ltd, Cancer Council Queensland PhD Scholarship and Professor William Burnett Research Fellowship from the Discipline of Surgery, School of Medicine, The University of Queensland.
DECLARATION OF INTEREST
None.
