Sunden 2010.

Methods	Study design: crossover RCT (3:1) Phase 1: blinded, randomised and placebo controlled crossover study Phase 2: blinded, observational and placebo controlled. Study duration: 4 years Duration of follow‐up: 24 months
Participants	Country: Sweden Setting: outpatient, single centre Adults with incomplete bladder emptying; 3 or more microbiologically proven UTI/year, 2 years prior to the study; optimal conservative measurements including clean intermittent catheterization to empty bladder; all 20 patients were recorded to have voiding dysfunction with at least 11 patients having neurological causes of bladder dysfunction. Number Phase 1: treatment group (13); control group (11) Phase 2: treatment group (11); control group (13) Mean age, range (years): 56.7, 32 to 84 Sex (M/F): 8/12 Comorbidities: none reported Exclusion criteria: upper UTI; renal deterioration; hydronephrosis; untreated bladder outflow obstruction; urinary calculi; immunosuppression; urological malignancies
Interventions	Treatment group Intravesical instillation: 30 mL of E. coli 83972 (105 CFU/mL); inoculation occurred once/d for 3 days Control group Intravesical instillation: 30 mL of saline; inoculation occurred once/d for 3 days
Outcomes	Self‐reported UTI symptoms Suprapubic pain/discomfort, malaise, fever, chills, urgency, leakage, abnormal urine, autonomic dysreflexia or spasticity Urine samples were taken for culture, the definition of uropathogenic growth was not defined
Notes	Patients who were unsuccessfully inoculated were not included in analysis. In Phase 1, median time to UTI were longer with successful E. coli 83972 colonisation than non E. coli colonisation ‐ 11.3 versus 5.7 months (P = 0.01). Phase 2 number of reported UTI was lower in patients with successful E. coli 83972 colonisation than those without (13 versus 35, P = 0.009) Funding: grants from Coloplast, Riksforbundet for Trafik och Polioskadade, Swedish Strategic Programme against antibiotic resistance, Region Skane FoU, Foundations of Gosta Jonsson, Hillevi Fries, Per‐‐Olof Strom and Greta Ekholm
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated randomisation list
Allocation concealment (selection bias)	Low risk	Wullt corresponded in 2015 that both active and placebo inoculum have the same appearance, a clear fluid that could not be differentiated by a non‐informed observer
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Patient were blinded as to the nature of inoculum; investigators were aware of whether patients were in intervention or control arm.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Wullt corresponded in 2015 that study team was unaware of urine culture results but were aware of patient's symptoms in view of outcome being self‐reported UTI symptoms. Wullt corresponded in 2015 that the investigator who surveyed the urine culture was aware of type of inoculum
Incomplete outcome data (attrition bias) All outcomes	High risk	Patients with unsuccessful colonisation post inoculation were excluded from analysis ‐ not fully intention to treat ‐ only analysed 20/26 patients. Wullt corresponded in 2015 that an intention‐to‐treat analysis was not conducted due to the small number of patients enrolled
Selective reporting (reporting bias)	High risk	Only reported that it is safe due to absence of pyelonephritis. In a later article published in 2014, the authors reported symptomatic UTI in 2 patients who were successfully inoculated with E. coli 83972
Other bias	Low risk	Study appears free of other biases

CFU ‐ colony‐forming units; DM ‐ diabetes mellitus; HPF ‐ high powered field; RCT ‐ randomised controlled trial; UTI ‐ urinary tract infection; VUR ‐ vesicoureteric reflux; WBC ‐ white blood cells