Urinary tract infection in primary care

Abstract

Recommendations for first line treatment should be informed by clinical and microbiological data

A recent prospective cohort study by McNulty and colleagues in the Journal of Antimicrobial Chemotherapy reports on 448 women with symptoms of uncomplicated urinary tract infection who were treated with trimethoprim in primary care.¹ The aim was to see whether women with infections resistant to trimethoprim had worse clinical outcomes. While the answer might seem intuitive, some of the findings were interesting. Pure bacterial culture was found in 317 women and the rate of resistance to trimethoprim was lower than expected from local laboratory resistance data derived from routinely collected specimens (13.9% v 24.5-27%). Predictably, antibiotic resistance was associated with longer median duration of symptoms (7 v 4 days; P<0.0002), higher frequency of subsequent prescription of antibiotics (36% v 4% in the first week; P<0.0001), and higher rates of reconsultation for treatment failure (39% v 6%; P<0.0001). While this sixfold relative difference in treatment failure rates is impressive, what is interesting from a primary care perspective is the low absolute reconsultation rate in the subsequent week in the resistant group (39%). In other words 61% of women with resistant organisms did not reconsult in the subsequent week because of treatment failure.

The treatment of uncomplicated urinary tract infection in primary care is usually empirical. The decision about which antibiotic to use may be influenced by both the practitioner's and the patient's previous experience, available data on antibiotic sensitivities, guidelines, and drug marketing.² General practitioners face two sometimes competing imperatives—the first to choose an effective treatment for the individual and the second to minimise resistance in the population by using antibiotics responsibly.

Data on local resistance are generally derived from routine clinical specimens being processed by community or hospital laboratories. Many sources of bias may exist in these data relating to referral patterns and pooling of results by organism rather than clinical condition. This bias results in overestimation of resistance rates in women with symptoms of uncomplicated urinary tract infections.^{3 4 5 6 7} The findings in this UK study concur with this—13.9% of patients in the study were resistant to trimethroprim compared with 24.5-27% in routinely collected specimens.¹

The authors call for more systematic and regular surveillance, which puts data into the prescribing context. Without such data, overestimations may influence prescribers to change their first line prescribing choices earlier than needed, especially if reinforced by drug companies promoting newer agents.

Ultimately, it is relief of symptoms that matters to patients, not microbiological eradication. We therefore need to use data on resistance with care when making decisions and developing guidelines for prescribing in primary care.

The British Society for Antimicrobial Therapy and US Clinical and Laboratory Standards Institute breakpoints (the minimum inhibitory concentration (MIC) standard in vitro that determines whether an organism is classified as “resistant” or “sensitive”) for determining resistance are best estimates of clinically important resistance. These are usually based on anticipated responses in bloodstream infections using pharmacodynamic, microbiological, and, where available, clinical response data. Many factors influence clinical outcome, including variable relations between concentrations of antibiotics in the blood and urine and patient characteristics.

The microbiologically determined resistance rates found by McNulty and colleagues might prompt a general practitioner to assume that a resistance rate of 13.9% equates to a similar treatment failure rate. If this were so, the number needed to investigate to change clinical outcome would be 10 (44/448). When reconsultation because of treatment failure due to resistance is the main outcome, the number needed to investigate rises to 26 to prevent reconsultation in the next week and 23 for the next month (20/448).¹ The same may be true for other antibiotics used to treat urinary tract infections.^{8 9} There may be wider lessons here about using intermediate outcome indicators like antibiotic resistance to guide prescribing decisions in general practice. Similarly, urine dipstick testing predicts significant bacteriuria but does not reliably predict response to antibiotic treatment.¹⁰ Taking a broader view, the limitations of risk factors as prognostic tools have recently been highlighted, as risk factors do not necessarily predict development of disease.¹¹

The authors claim that their data support trimethoprim as an appropriate first line agent for uncomplicated urinary tract infection in their region. It is clinically effective, relatively safe, and inexpensive. Trimethoprim is alone in its class, which reduces the likelihood of resistance selection to other, newer antibiotics, and it is rarely, if ever, used for more serious infections. We agree with the authors' conclusions that the decision to switch to a second antibiotic should be made on clinical grounds rather than on microbiological grounds—that is, failure of symptoms to resolve after four days of treatment. With a clinical failure rate of 17/448, general practitioners can confidently tell patients that most women's symptoms will resolve quickly, and that they should return if symptoms are not improving by four days—sooner if symptoms worsen. Laboratory investigation seems warranted only if initial treatment fails.

The relation between laboratory determined and clinical resistance will not be constant, or necessarily generalisable from this study. It follows that choice of first line treatment should be informed by periodic and systematic community surveillance using clinical and laboratory defined outcomes. This is likely to be cheaper overall than routinely ordering pretreatment investigations that are unlikely to be helpful for empirical prescription and may lead to unnecessary second prescriptions.

Many questions still need to be answered. Just as the relation between prescribing patterns, resistance, and clinical outcomes is complex,¹² the association between detectable infection and response to antibiotics is not linear.¹⁰ Ironically, rigid prescribing guidelines for first line treatment may be less helpful in containing antibiotic resistance. It has been suggested that (rational) diversity of first line agent may dilute selection pressure, and further evaluation of a variety of strategies is needed.¹² We do not know why some people with symptoms respond to antibiotics faster than to placebo when they do not have infection by any accepted definition, while others with sensitive organisms fail to respond to antibiotics.