We identified only three small European trials, with follow-up ranging from 6 to 15 months, in older women.
We did not identify any trials in older men nor any trials in frail care home residents.
Recurrent urinary tract infection is one of the most common reasons for long-term antibiotic use in the frail elderly. We systematically reviewed trial evidence to address clinical uncertainties around this practice.
There are several important clinical uncertainties relating to long-term antibiotic use in older adults with recurrent UTI, including effect on frequency of infective episodes, optimal duration of prophylaxis, adverse effects, risk of relapse following cessation of prophylaxis and effect on urinary antibiotic resistance. We therefore systematically reviewed randomised controlled trials comparing long-term antibiotic prophylaxis with placebo or non-antibiotic therapy for preventing further episodes of UTI in older people. Our main objective was to quantify the benefits and harms of long-term antibiotics for older adults, to better inform patients and clinicians during clinical decision making.
Previous meta-analyses showed antibiotic prophylaxis conferred a relative risk reduction of 79% in the proportion of women experiencing a microbiologically confirmed UTI, compared with placebo. 8 However, these analyses included data from mostly small trials of younger women without comorbidities. There is uncertainty around the generalisability of these findings to older adults.
UTIs, and consequently recurrent UTIs, are overdiagnosed in older people. 4 5 Therefore, antibiotic prophylaxis may actually be prescribed for symptoms that represent bladder dysfunction or localised vaginal symptoms rather than true UTI, and thus will not confer the intended benefit. Multimorbidity, frailty and polypharmacy are more common in older people and are contributory factors for potential harms such as those related to drug interactions. For example, older adults coprescribed renin–angiotensin system inhibitors and trimethoprim-containing antibiotics were shown to be at increased risk of hyperkalaemia-related hospitalisation 6 and sudden death. 7
Older men and women are commonly prescribed long-term antibiotics to prevent recurrent urinary tract infection (UTI). 1 2 Antibiotic use is a key driver of antibiotic resistance. 3 Therefore, antibiotic use must be justified by robust evidence, where the estimated benefit outweighs estimated harm.
Outcomes measured in only one trial were reported narratively. Outcomes measured in more than one trial were synthesised quantitatively. We estimated between trial heterogeneity using the I 2 statistic 11 and used random effects meta-analyses to estimate pooled risk ratios and 95% CIs. 12 We undertook sensitivity analyses to examine treatment effects according to study quality and assessed the impact of including data from a potentially eligible trial where the study author did not reply to our request for data on older participants.
One reviewer (HA) extracted study characteristics (setting, participants, intervention, control, funding source) and outcome data from included trials. We contacted two authors for subgroup data on postmenopausal women. One author replied and provided relevant outcome data. Two reviewers (HA and SP) independently assessed the risk of bias of the included studies using the Cochrane Collaboration’s risk of bias tool. 10 Disagreements were resolved through discussion. We used RevMan V.5.3 to meta-analyse the data and generate forest plots.
Our primary outcome was the number of UTI recurrences per patient-year during the prophylaxis period, defined microbiologically (>100 000 colony-forming units of bacteria/mL of urine) and/or clinically (for example, dysuria, polyuria, loin pain, fever) or other measure of change in the frequency of UTI events during prophylaxis. We also aimed to assess the proportion of patients with severe (requiring withdrawal of treatment) and mild (not requiring withdrawal of treatment) adverse effects. Secondary outcomes included the proportion of patients who experienced at least one recurrence after the prophylaxis period, time to first recurrence, proportion of patients with antibiotic resistant micro-organisms in future urine samples and quality of life.
We excluded studies evaluating the effect of prophylactic antibiotics in specific situations, for example, post catheterisation, postsurgery, in patients with spinal injuries or in those with structural renal tract abnormalities.
We included studies recruiting adults of all ages and screened relevant results to assess whether reported data allowed estimates of effect size in our specified population of older adults. For data not presented in this format, we contacted authors if the study was published in the last 10 years and if the mean or median age in any arm was greater than 50 years.
We included only randomised controlled trials published in full (ie, not abstracts) in English, comparing the effect of long-term antibiotics versus placebo or non-antibiotic interventions on the rate of UTI in older adults with recurrent UTI. We defined ‘long-term antibiotics’ as daily antibiotic dosing for at least 6 months, ‘older adults’ as women who were postmenopausal or over the age of 65 and men aged over 65 and ‘recurrent UTI’ as self-reported or clinically recorded history of two or more UTIs in 6 months or three or more in 12 months.
One author (HA) conducted the first screening of potentially relevant records based on titles and abstracts, and two authors (HA and FD) independently performed the final selection of included trials based on full-text evaluation. Reference lists of included studies and relevant systematic reviews were screened for further potentially relevant studies. Disagreements between the two reviewers were resolved through discussion.
We systematically searched Medline, Embase, CINAHL and the Cochrane Central Register of Controlled Trials from inception to March 2016 for English language randomised controlled trials. Our search strategy consisted of keywords and medical subject headings terms for urinary tract infection and randomised trials (see online supplementary appendix 1 ).
We conducted a systematic review following guidance from the Cochrane Handbook for Systematic Reviews of Interventions for conduct and PRISMA guidelines (see online supplementary file PRISMA checklist ) for reporting. 9 The review protocol was prospectively registered on the International Prospective Register of Systematic Reviews (PROSPERO :%20http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015016628 ; registration number: PROSPERO 2015:CRD42015016628).
From 6645 records, we identified 53 studies for full-text review (see online supplementary appendix 2). Four studies were eligible for inclusion.13–16 Two studies recruited only postmenopausal women.15 16 Two studies recruited women of all ages but the median age was >50 years.13 14 For these studies, we contacted authors requesting data for postmenopausal women, or if menopausal status not ascertained, for women aged over 65. We received data from one author and hence included three trials consisting of 534 postmenopausal women in our review (table 1).14–16 We did not identify any studies that included older men.
Characteristics of included studies
Trials were conducted in community and outpatient settings in Israel, the Netherlands and Croatia. Only one trial included individuals with diabetes16 and only one trial included individuals with renal impairment.14 Intervention arms consisted of 6 to 12 months of antibiotic therapy. Control arms consisted of non-antibiotic prophylaxis with vaginal oestrogen pessaries,15 oral lactobacilli capsules16 and D-mannose powder.14 One trial reported the number of UTI recurrences per patient year during the prophylaxis period.16 All trials reported the number of women experiencing a UTI during the prophylaxis period and frequency of adverse events. Only one trial assessed recurrence of UTI after the prophylaxis perio.16 One trial assessed effect on urinary and faecal bacterial resistance.16
Risk of bias
Figure 1 summarises the risk of bias assessment. Allocation and randomisation details were poorly reported in two trials.14 15 One trial was assessed as high risk for performance and detection bias; trial arms consisted of an oral antibiotic capsule or D-mannose powder diluted in 200 mL water or no treatment with no use of placebo and did not report on blinding of outcome assessors.14 Only one trial reported a sample size calculation.14 Overall, one trial was judged to be low risk of bias16 and two trials unclear risk due to limited reporting of methods.14 15
Summary of risk of bias assessment.
Effect of long-term antibiotics on recurrent UTI
Compared with a capsule of lactobacilli, prophylaxis with 480 mg of trimethoprim–sulfamethoxazole for 12 months led to fewer microbiologically confirmed UTI episodes per patient-year (mean number of episodes per year=1.2 vs 1.8, mean difference 0.6, 95% CI 0.0 to 1.4, p=0.02). Prophylaxis with trimethoprim–sulfamethoxazole also led to less women experiencing a microbiologically confirmed UTI during prophylaxis (49.4% vs 62.9%; RR 0.79, 95% CI 0.63 to 1.0) and an increase in time to first UTI (6 months versus 3 months; log-rank p=0.02). There was no difference between arms in the mean number of microbiologically confirmed UTI episodes 3 months after cessation of prophylaxis (mean number of episodes=0.1 vs 0.2, mean difference 0.0, 95% CI −0.1 to 0.3, p=0.64).16
Compared with vaginal oestrogen pessaries, prophylaxis with 100 mg of nitrofurantoin for 9 months led to fewer women experiencing a UTI during prophylaxis (42.3% vs 64.6%; RR 0.65, 95% CI 0.8 to 0.90) and a lower mean number of UTIs per woman (0.6 episodes per woman vs 1.6 episodes per woman).15
Compared with D-mannose powder prophylaxis with 50 mg of nitrofurantoin for 6 months led to more postmenopausal women experiencing a UTI during prophylaxis (24% vs 19%, RR 1.24, 95% CI 0.57 to 2.69).14
Random effects meta-analysis (figure 2) shows long-term antibiotic therapy reduces the risk of a woman experiencing a UTI during the prophylaxis period (pooled RR 0.76; 95% CI 0.61 to 0.95) with about eight post-menopausal women needing treatment with long-term antibiotics to prevent one woman experiencing a UTI during the prophylaxis period (number needed to treat (NNT)=8.5).
Forest plot showing results of meta-analysis for proportion of women experiencing a UTI during the prophylaxis period. UTI, urinary tract infection.
Commonly reported side effects across the three trials included skin rash, gastrointestinal disturbance and vaginal symptoms. There were no statistically significant difference between risk of adverse events between trimethoprim–sulfamethoxazole and lactobacilli,16 or between nitrofurantoin and vaginal oestrogens.15 Risk of side effects with D-mannose powder was significantly lower than with nitrofurantoin (RR 0.28; 95% CI 0.13 to 0.57).14 Overall, absolute numbers of serious adverse events or events resulting in treatment withdrawal were small.
We had data on mild adverse events (not resulting in treatment withdrawal) for all three trials. There was marked heterogeneity between trials for adverse events (I2=86%).
Meta-analyses showed no statistically significant difference between antibiotics and control for overall risk of mild adverse events (pooled RR 1.52; 95% CI 0.76 to 3.03) (figure 3).
Forest plot showing results of meta-analysis for proportion of women experiencing mild side effect (treatment not withdrawn) during the prophylaxis period.
We extracted data for serious adverse events (resulting in treatment withdrawal) for two trials. Meta-analyses showed no statistically significant difference between antibiotics and control for overall risk of serious adverse events (pooled RR 0.90; 95% CI 0.31 to 2.66; figure 4).
Forest plot showing results of meta-analysis for proportion of women experiencing a serious side effect (resulting in treatment withdrawal) during the prophylaxis period.
Effect of long-term antibiotic therapy on bacterial resistance
Compared with lactobacilli, women receiving 12 months prophylaxis with trimethoprim–sulfamethoxazole showed dramatic increases in the proportion of antibiotic resistant bacteria isolated from urine and faeces. For example, 20%–40% of urinary and faecal E coli isolates were resistant to trimethoprim–sulfamethoxazole, trimethoprim and amoxicillin at baseline, increasing to 80%–95% after 1 month of treatment. Over the 15 month follow-up period, resistance levels decreased following cessation of prophylaxis but remained above baseline levels.16
We assessed the impact of removing the study at high risk of bias on effect size and direction.14 Removal made little difference to the meta-analysis for proportion of women experiencing a UTI during the prophylaxis period (pooled RR 0.74; 95% CI 0.61 to 0.89). Removal did impact on the meta-analysis for proportion of women experiencing mild side effects during the prophylaxis period but overall difference between antibiotics and placebo did not reach statistical significance (pooled RR 0.99, 95% CI 0.82 to 1.20).
We also pooled aggregate data from another potentially relevant study where authors did not respond to our request for data regarding postmenopausal women/women over 65.13 This study compared 500 mg of cranberry extract to 100 mg trimethoprim taken at night for 6 months. However, adding aggregate data for the whole study population (women aged 45 and above) to our meta-analysis for the proportion of women experiencing a UTI during the prophylaxis period made little difference to risk estimates (pooled RR 0.74; 95% CI 0.61 to 0.90).