Results of the search

We identified 389 records. Following assessment of titles and abstracts, 28 full‐text records were screened. Of these, nine studies (14 records) were included and eight studies (10 records) were excluded. Two ongoing studies were identified (NCT00781625; ProSCIUTTU Study 2014), one study is awaiting translation (Skerk 2010), and one study was identified prior to publication ( Reid 1995). These four studies and will be assessed in a future update of this review (Figure 1).

Figure 1

---

---

Study flow diagram

Included studies

We included nine studies in this review (Baerheim 1994; Czaja 2007; Ferrara 2009; Kontiokari 2001; Lee 2007a; NAPRUTI Study II 2006; Reid 1992; Reid 2003; Stapleton 2011).

Six studies compared probiotics with placebo (Baerheim 1994; Czaja 2007; Reid 1992; Stapleton 2011) or no comparator (Ferrara 2009; Kontiokari 2001); two studies compared probiotics with antibiotic prophylaxis in patients with UTI (one in adults (NAPRUTI Study II 2006) and one in children with VUR (Lee 2007a)); and one study compared probiotics with placebo in healthy women (Reid 2003).

Excluded studies

Seven studies were excluded (Characteristics of excluded studies). Dani 2002 was conducted in a neonatal population; Colodner 2003 did not include an arm where participants did not receive a probiotic; Molander 1990 did not include investigation of a probiotic; and NCT00900653 studied a combination of probiotic and hormonal therapies. Manley 2007 and Pushkarev 2005 studied treatment rather than prophylaxis; Ranganathan 2009 enrolled people with chronic kidney disease (CKD) to determine if probiotics exerted renoprotective effects and reduced uraemia symptoms.

Allocation

Three studies (Ferrara 2009; Kontiokari 2001; Stapleton 2011) reported on adequate methods of sequence generation and the other six studies (Baerheim 1994; Czaja 2007; Lee 2007a; NAPRUTI Study II 2006; Reid 1992; Reid 2003) did not describe sequence generation methods.

Four studies reported adequate allocation concealment (Czaja 2007; NAPRUTI Study II 2006; Reid 1992; Stapleton 2011). Allocation concealment was unclear for the remaining five studies.

Blinding

Two included studies were adequately described as double blind (NAPRUTI Study II 2006; Stapleton 2011), two were open label (Ferrara 2009; Kontiokari 2001) and there were insufficient data to assess blinding for the remaining studies; blinding was either not reported or lacked sufficient detail to determine who was blinded.

Incomplete outcome data

Satisfactory explanation was provided for changes in the number of participants for only one of the studies assessed (Czaja 2007). Reid 2003 had a significant proportion of their small study population excluded from analysis; attrition was significant and not explained satisfactorily in three studies (Kontiokari 2001; NAPRUTI Study II 2006, Stapleton 2011). Remaining studies lacked sufficient information to determine if attrition was likely to result in significant bias. Due to this incomplete follow‐up, worst case scenarios were undertaken for both the probiotics in comparison to placebo analysis and the probiotics in comparison to antibiotics analysis.

Selective reporting

Many of the included studies either did not report secondary outcomes, or lacked sufficient detail in reporting secondary outcomes; many of our prespecified secondary outcomes were not addressed, including all‐cause mortality and number with at least one confirmed case of bacteraemia or fungaemia.

We searched for published protocols for the included studies. Protocols were found for NAPRUTI Study II 2006 and Stapleton 2011. The outcomes reported in Stapleton 2011 aligned completely with the published protocol. The published report of the NAPRUTI Study II 2006 included several outcomes that were not prespecified in the protocol: mean number of antibiotic prescriptions for UTI treatment; and a subgroup analysis of mean number of clinical recurrences in women with complicated versus uncomplicated UTI.

Other potential sources of bias

Several studies were funded by manufacturing companies (Czaja 2007; Reid 1992), and one had an issue with supply that resulted in treatment duration inequality between the study arms (Kontiokari 2001). UTI definitions were fairly consistent among studies, although microbiological criteria ranged from at least 10³ CFU/mL to 10⁵ CFU/mL and clinical criteria were more stringent in some studies compared with others.

Primary outcome

The primary outcome was to assess numbers of participants with at least one symptomatic bacterial UTI in each group (as confirmed by a catheter specimen of urine, midstream urine specimen if possible, or a clean catch specimen, and defined as > 10⁵ CFU/mL or as defined by triallists).

Analyses were conducted according to probiotics in women and probiotics in children with VUR. Placebo‐controlled studies were subdivided into studies that enrolled women and children who had recently been treated with antibiotics for UTI (Ferrara 2009; Kontiokari 2001; Reid 1992; Stapleton 2011) and studies enrolling participants who had previous UTIs (Baerheim 1994; Czaja 2007). We decided to first determine if probiotics exerted a positive effect versus placebo before analysing studies versus active comparators.

A meta‐analysis of six studies that involved 352 randomised women and children demonstrated no significant reduction in the risk of recurrent symptomatic bacterial UTI between probiotics and placebo (Analysis 1.1 (6 studies, 352 participants): RR 0.82, 95% CI 0.60 to 1.12; I² = 23%), heterogeneity was low. The confidence interval for all studies suggests a range that includes a 15.8 % absolute decrease to a 4.7% absolute increase in the risk of recurrent bacterial UTI with probiotics versus placebo given that the placebo rate of recurrence was 39.5% over 8 to 52 weeks.

NAPRUTI Study II 2006 reported no significant difference in the rate of recurrent symptomatic bacterial UTI in women between probiotics and antibiotics (Analysis 2.1 (1 study, 223 women): RR 1.12, 95% CI 0.95 to 1.33).

Lee 2007a included children with VUR. There was no significant difference in the rate of recurrent symptomatic bacterial UTI between probiotics and antibiotics (Analysis 3.1 (1 study, 96 children): RR 0.54, 95% CI 0.24 to 1.23).

We analysed six month recurrent UTI data from Kontiokari 2001. However, the authors also reported recurrent UTI at 12 months; a sensitivity analysis demonstrated that the results of the meta‐analysis did not change meaningfully if 12 month data were used. Three studies did not report on this outcome and this may have improved the precision of the effect size if the data were available. Stapleton 2011 reported 17 and 13 recurrent symptomatic bacterial UTIs for probiotics and placebo respectively. However, correspondence with the lead author indicated that two women in each arm of the study had symptomatic UTI but these were not confirmed with positive cultures; therefore, these events were excluded. We included these unconfirmed UTI in our analysis. In a sensitivity analysis, the results of the meta‐analysis did not change meaningfully if only culture‐confirmed UTI data were used.

Secondary outcomes

None of the included studies reported numbers of participants with at least one asymptomatic bacterial UTI, all‐cause mortality or those with at least one confirmed case of bacteraemia or fungaemia. The only secondary outcomes of interest reported by the included studies were withdrawal due to adverse events, total adverse events and numbers of participants with at least one non‐fatal serious adverse event.

Withdrawal due to adverse events was reported by NAPRUTI Study II 2006 and Stapleton 2011. NAPRUTI Study II 2006 reported that withdrawals due to adverse events occurred in six probiotic participants (5.2%) versus 15 antibiotic participants (12.2%). Stapleton 2011 noted that one placebo group participant discontinued treatment due to an adverse event.

Ferrara 2009, Kontiokari 2001, and Reid 1992 did not report adverse events. NAPRUTI Study II 2006 and Stapleton 2011 reported participants who experienced at least one adverse event but data were not meta‐analysed because participants in the NAPRUTI Study II 2006 control group received an antibiotic, and Stapleton 2011 provided a placebo to control group participants. NAPRUTI Study II 2006 reported that 66 probiotic (57.4%) and 72 antibiotic participants (58.5%) respectively experienced at least one adverse event (the most commonly reported adverse effects in both groups were diarrhoea, nausea, vomiting, constipation and vaginal symptoms). Reid 2003 identified adverse events through a questionnaire that was sent to patients. They reported that no probiotic patients reported an adverse event however it is unclear how many comparator group patients experienced at least one adverse event; they only reported that 2 comparator patients reported yeast infections.

Stapleton 2011 reported that 28 probiotic (56%) and 25 placebo participants (50%) experienced at least one adverse event (the most common were vaginal discharge, itching and moderate abdominal discomfort).

Baerheim 1994 stated that treatment was well tolerated in both groups; four participants from the probiotic arm and one from placebo complained of discharge, with no other adverse effects noted.

Czaja 2007 documented a range of self‐reported adverse effects including abnormal vaginal discharge, external genital irritation, vaginal candidiasis, vaginal odour, abdominal pain and dysuria. Abnormal vaginal discharge occurred in about half of all participants, but the overall frequency of adverse effects was low.

NAPRUTI Study II 2006 and Czaja 2007 reported serious adverse events. In NAPRUTI Study II 2006, no significant difference in serious adverse events was noted; 17 probiotic (14.8%) and 14 antibiotic participants (11.4%) experienced at least one serious adverse event.

Subgroup analyses

In a post‐hoc subgroup analysis, there was no significant difference in recurrence of UTI between the subgroups of women without a UTI prior to enrolment compared (Analysis 1.1.1 (2 studies, 77 participants): RR 1.12, 95% CI 0.65 to 1.93; I² = 0%) with those with UTI being treated with antimicrobials at enrolment (Analysis 1.1.2 (4 studies, 275 participants): RR 0.74, 95% CI 0.52 to 1.05; I² = 16%). The overall pooled estimate for all studies in both subgroups was not significantly different from the pooled estimate of each subgroup (Test for subgroup differences: Chi² = 1.57, df = 1 (P = 0.21), I² = 36.5%).

Sensitivity analyses

Sensitivity analyses were planned to determine if treatment effects on recurrent symptomatic bacterial UTI rates differed in studies based on a number of variables. Removal of studies that were open label or that had unclear allocation concealment did not change the effect of probiotics on recurrent symptomatic bacterial UTI however only few studies had unclear allocation concealment (Baerheim 1994; Ferrara 2009; Kontiokari 2001) or were open label (Ferrara 2009; Kontiokari 2001) and it was felt that sensitivity analyses would not be informative and maybe misleading.

Sensitivity analysis was also conducted for symptomatic bacterial UTI involving Czaja 2007, NAPRUTI Study II 2006 and Stapleton 2011 as these three studies suggested that not all randomised patients were included in the UTI analysis; hence we wanted to see the impact of imputing data for a worst case scenario. In this analysis all missing patients in one group were assumed to have had UTI, and those missing from the other group were assumed to not have had UTI. For the comparison of probiotics versus placebo, the effect did not change using a worst case scenario for probiotics (Analysis 1.2: RR 0.94, 95% CI 0.63 to 1.39; I² = 49%). For the comparison of probiotics versus antibiotics, a worst case scenario for antibiotics of the NAPRUTI Study II 2006 now demonstrated that fewer probiotic patients experienced a recurrent symptomatic bacterial UTI versus antibiotics (25 antibiotic patients, Analysis 2.3; RR 0.83, 95% CI 0.74 to 0.94). When the worst case scenario analysis for probiotics was conducted, more probiotic patients had recurrent symptomatic bacterial UTI (5 probiotic patients, Analysis 2.2; RR 1.19, 95% CI 1.01 to 1.40).