Background

Symptomatic urinary tract infection (UTI) in women is a common condition in general practice, and every day general practitioners or their staff instruct women in delivering urine samples for examination [1]. The main concern when sampling urine is that inadequate handling may increase the risk of contamination in turn leading to overdiagnosis and overtreatment of UTI. Sterile collection of urine samples can be performed using suprapubic puncture or urethral catheterization and use of these collection techniques could possibly reduce contamination and thereby overdiagnosis and overtreatment. However, in a primary care setting these methods are considered obsolete today due to the associated discomfort for the patient and a minor risk of iatrogenic infection and other complications. Current methods include i) mid-stream-clean-catch technique (MSCC) where the patient is instructed to clean the labia before voiding using tap water, soap or disinfectants, ii) mid-stream urine (MSU) without prior cleaning, iii) random samples delivered without instruction or iv) home-voided samples with or without standardized transport media. These sampling techniques are mostly based on tradition or expert opinion and ease-of-use for patient and doctor rather than stringent evidence. A study from 2000 conducted in primary care found no evidence that sampling technique affected contamination rate or infection rate in urine samples [2], but new evidence within this area is often questioned and debated [3–5]. Since sampling techniques (MSCC, MSU, random samples and home voiding) differ extensively in preparation time and discomfort, ease-of-use for doctors as well as their patients, it is relevant to review their diagnostic yield. The aim of this study was to conduct a systematic review to determine the accuracy of urine culture from different sampling techniques in symptomatic patients in primary care.

Method

Literature search

We searched Medline and Embase for clinical studies conducted in primary care published before May 2015 in English, Swedish, Danish or Norwegian. Combinations of the words “urinary tract infection”, cystitis, bacteriuria, urine, specimen, handling, urinalysis, collection, midstream and” clean catch” were used. To identify more studies from before 1970, a slightly different search-string was used for the older studies in Medline. The literature search and inclusion of studies was performed by AH. The full search strings can be seen in Appendix A.

Inclusion criteria

Clinical studies randomizing or using a paired design to compare the result of urine culture obtained with two or more collection techniques in adult, self-helped, non-pregnant (and not post-partum) women with symptoms of UTI in primary care (general practice, outpatients clinics or comparable settings). We did not discriminate between complicated and uncomplicated cases of UTI.

Exclusion criteria

  • Studies investigating mainly patients who were not self-helped, were asymptomatic, pregnant, children or men (wrong group)

  • Studies conducted in the secondary sector (wrong setting)

  • Studies using other modalities than culture as reference (wrong gold standard)

  • Studies where data for the selected outcome was not available (missing data)

  • Studies using a different design than described in the inclusion criteria (wrong design)

The references of included studies were screened and experts in the field were contacted to provide additional literature.

Data extraction

Data from included studies were entered into a data-form with information on setting, number of patients, age, inclusion- and exclusion-criteria for the study, reference and index text, the assigned cut off for infection vs. contamination, the bacteria identified and study design. Data on absolute numbers of infected urine samples, true and false positives and negatives or predictive values of one sampling method versus another were likewise extracted from the included studies. If these measures were not directly provided in the article, we calculated them if possible. Selected outcomes were dichotomized for the planned analyses as negative/positive culture. Culture results presented as equivocal and contaminated were grouped with the negative results. Data from the relevant patients were extracted when studies also included patients covered by the exclusion criteria. Data extraction was done by both authors and discrepancies were discussed and corrected. When data was not available or incomplete we referred from contacting authors, as most studies were more than 10 years old.

Definition of reference standard

Assuming an increasing contamination rate in the order of: 1) Suprapubic puncture, 2) urethral catheterization samples, 3) MSCC, 4) MSU, 5) Random samples, 6) Home-voided urine, the least contaminated was used as reference and the most contaminated as index test. For example, if a study investigated both MSCC and random urine sampling in a paired design, MSCC was used as reference standard and random samples as index test.

Study designs

This review included both paired studies and randomized controlled trials (RCT). RCTs were analysed separately.

Quality assessment

The included studies were evaluated using QUADAS-2 for assessment of diagnostic accuracy studies [6]. No study was excluded based on low quality according to this tool. Both studies using paired samples and randomized controlled trials were assessed with QUADAS-2.

Data analysis

The specified dichotomized outcomes were used to calculate predictive values, sensitivity and specificity in paired studies. The generated sensitivity and specificity values were used to create forest plots on the diagnostic accuracy. Diagnostic accuracy plots were performed using Review Manager (RevMan) Version 5.3. Copenhagen: The Nordic Cochrane Centre, the Cochrane Collaboration, 2014.

Results

Literature search

The initial search resulted in 570 titles in Medline and 749 titles in Embase. After review of titles, abstracts and articles we included seven full text articles presenting results from seven studies investigating urine sampling technique in 1062 non-pregnant women with symptoms of UTI in primary care. A flow diagram of the literature search and review of titles, abstracts, and articles is shown in Fig. 1. Two of the studies were from general practice while the remaining five were from outpatient clinics or student clinics. The included studies are shown in Table 1. A list of excluded studies is provided in Appendix B.

Fig. 1
figure 1

Short legend: data collection

Full size image
Table 1 Characteristics of included studies
Full size table

Quality of included studies according to Quadas-2

The quality of the included studies is summarized in Table 2. Generally the studies were judged to be of moderate to high risk of bias. No study was considered having low risk of bias. The most common error was lack of blinding of the interpreter to the results of the index and reference tests or lack of reporting of blinding. The applicability of the studies was not regarded a general concern. The full quality assessment is described in Additional file 1.

Table 2 Quality of included studies assessed using Qaudas-2
Full size table

Data from included studies

Paired design studies

Four studies used a paired design to compare MSCC urine samples to samples obtained with urethral catheterization or suprapubic puncture (n = 589) [7–10]. Urethral catheterization and suprapubic puncture are essentially sterile techniques and served as reference e.g. gold standard. Two of these studies applied ≥10 cfu/ml as the cut-off for infection in both index- and reference-test, one study used a cut-off of ≥ 105 cfu/ml and one reported absolute counts for both index- and reference-test (Table 1). The positive predictive value (PPV) of a MSCC sample varied according to the chosen cut-off for infection: cutoff: ≥10 cfu/ml 0.79 (0.71-0.86); cutoff: ≥ 105 cfu/ml 0.95 (0.83-0.99). The negative predictive value of a MSCC was close to 1 in all four studies. The accuracy found in the four studies is shown in Table 3. The achieved specificity was influenced by the selected cut-off levels, with higher thresholds corresponding to increasing specificity. We did not perform a meta-analysis or calculate heterogeneity as the applied cut-offs varied considerable thus impeding a meaningful pooling of the results.

Table 3 MCSS vs. sterile samples
Full size table

One study investigated home-voided samples against MSCC taken in general practice [11]. This study found a high PPV of home-voided samples of 0.92 (0.81-0.98), but a lower NPV of 0.71 (0.48-0.88). The results of this study are shown in Table 4.

Table 4 Home-voided samples vs. MSCC
Full size table

The studies by Stamm and Mabeck reported absolute counts of colony-forming units in the voided urine samples and this allowed us to investigate the current cut-off for primary uropathogens of 103 cfu/ml in voided urine samples against 10 cfu/ml in suprapubic puncture [12]. Using these current cut-offs we calculated the sensitivity of MSCC to be 0.81 (0.71-0.88) in the study by Stamm and 0.96 (0.85-0.99) in the study by Mabeck. Corresponding specificities were 0.90 (0.82-0.95) in the study by Stamm and 0.59 (0.43-0.73) in the study by Mabeck.

Randomized controlled trials

Two randomized controlled trials were identified comparing MSU or random samples to MSCC with infection rate and contamination rate in the randomization-groups as their primary outcomes (number of patients = 400) [2, 13]. Because of the randomized design, accuracy could not be calculated from these studies. The studies are shown in Table 5. None of the studies found significant differences in infection rate or contamination rate between sampling techniques.

Table 5 Randomized controlled trials and infection rates
Full size table

Discussion

This diagnostic accuracy review is the first to assess the available evidence from different urine sampling techniques on symptomatic patients with suspected UTI in primary care. Overall, we did not find consistent evidence to suggest important differences in diagnostic accuracy among the included urine sampling techniques (MSCC, MSU or random voiding). The slightly lower specificity of voided samples compared to invasive sampling techniques (suprapubic puncture and catheter) will cause 5–10 % of healthy patients to be overdiagnosed. This does not, in our opinion, outweigh the discomfort and risk of complications associated with sterile techniques. The quality of the studies was moderate and substantial heterogeneity was present between study designs and applied diagnostic cut-offs. With the available evidence, each general practitioner can choose freely the sampling technique most appropriate for his or her practice and patients.

The current review included two studies from general practice and 5 from outpatient clinics or student clinics. Participants were symptomatic patients under investigation for urinary tract infection. We have no reason to suspect the included patients differ from the average UTI patient in primary care. Thus we believe the results can be considered applicable to most primary care settings including general practice.

The included methods of urine sampling included, the different cut-offs for infection applied and the time span between studies of up to 50 years does however suggest that the overall results regarding their diagnostic accuracy should be considered with caution.

The current consensus regarding a cut-off for infection (eg. 103 cfu/ml for primary uropathogens) was not directly assessed in any of the studies, but we calculated the sensitivity and specificity based on the two studies by Mabeck and Stamm. While the sensitivity was above 0.80 in both studies, the specificity differed between studies and was low (0.59) in the study by Mabeck. However, this could be a chance finding and caution should be excised when interpreting these results as they are based on few older studies and we do not know if this result would still apply today with current microbiological procedures. Furthermore, current cut-offs are based on microbiological assessments and have, to our knowledge, never been validated in relation to patient-relevant outcomes like cure-rate or impact on daily activities. The development of such patient-centred outcomes may be more applicable to a general practice setting.

The European urine analysis guideline recommends a MSCC without detergents [12]. However, this guideline is based on studies including pregnant, asymptomatic as well as hospitalized patients and their conclusions do not necessarily apply to the average patient in general practice. Studies based in secondary care have found varying accuracy of voided urine samples depending on their patient group, design and gold standard [14–18]. However, studies investigating symptomatic, otherwise healthy women seem to essentially reproduce our findings [19, 20].

Conclusions

The present review does not present evidence to suggest one urine sampling technique over another according to diagnostic performance; rather this should at present depend on ease of use and convenience for patients and practices. This lack of evidence is in part due to few available studies and further testing on current diagnostic cut-offs as well as new patient-centred outcomes is warranted.

Abbreviations

FN, false negatives; FP, false positives; MSCC, mid-stream-clean-catch technique; MSU, mid-stream urine; NPV, negative predictive value; PPV, positive predictive value; RCT, randomized controlled trials; SEN, sensitivity; SPE, specificity; TN, true negatives; TP, true positives; UTI, urinary tract infection.