• routine laboratory data
• antibiotic resistance
• empirical prescribing
• urinary tract infection

Smellie and colleagues have noted large and significant differences in rates of submission of samples for microbiological testing between practices in the south west and north east regions.¹ They think that these data indicate that some practices are more selective in the use of the laboratory and speculate that routine laboratory generated antibiotic surveillance data will thus tend to include more complicated cases and overestimate antibiotic resistance. The authors therefore question the validity of using routine laboratory data on antibiotic resistance for primary care based therapeutic guidelines for empirical prescribing, and recommend an enhanced surveillance programme with a standardised approach to testing as a better way of informing such guidelines. We have recently completed such a programme for urinary tract infection (UTI) and can therefore comment on these suggestions.

The susceptibility data for uropathogens recovered from routine urine samples received from the 80 practices served by Cambridge Microbiology and Public Health Laboratory were compared with the results from a sentinel group of five practices in the same locality, which agreed to submit urine samples on all patients presenting with a clinical diagnosis of UTI within a three month period during 2002.

In total, 967 urine samples were received from the sentinel practice group and 18 892 from the general practice group. Bacteria were recovered from 269 and 4449 samples from the two groups, respectively. Overall, 89% were Gram negative bacilli and the numbers of these were large enough to permit meaningful comparison and statistical evaluation. There were no significant differences in recovery of any species between the two groups, indicating that, in contrast to the assertion of Smellie et al, any bias by general practitioners to send in samples with complicated infections involving more resistant organisms, such as Pseudomonas spp, did not result in overestimating the extent of antibiotic resistance in our population.

Resistance rates to cefalexin, norfloxacin, and gentamicin were marginally lower (2.5% v 5.2%, 2.1% v 4.4%, and 0.8% v 1.3%, respectively) and resistance to ampicillin, trimethoprim, and co-amoxiclav was slightly higher (46.0% v 45.1%, 22.8% v 19.8%, 18.1% v 11.7%, respectively) in the sentinel practice group than in the general practice group. However, only the resistance rate for co-amoxiclav was significantly different (p  =  0.03 by two tailed χ² test).

Two other studies of enhanced surveillance programmes have been published, also investigating UTI. Baerheim et al reported a study on female patients with UTI in general practice in Norway, comparing resistance rates for bacteria recovered from unselected (sentinel) patients with those from whom urine was sent routinely, using a panel of six antibiotics.² Uropathogens from the sentinel group were 3.4–8.4% less resistant to the antibiotics tested, but the results were significant only for nitrofurantoin (3.0% resistant in sentinel patients compared with 9.7% resistant in routine patients).

Richards reported a study similar to our own over a 12 month period involving three sentinel general practices in Norfolk, UK.³ Sentinel practice resistance rates were the same, or very slightly higher for co-amoxiclav, ciprofloxacin, cefradine, and gentamicin and slightly lower for ampicillin, cefuroxime, and trimethoprim. The difference was largest with trimethoprim (18% resistant in sentinel practice patients versus 22% resistant in all other practices), and only this difference achieved significance. We did not duplicate their findings; indeed, the pattern for lesser or greater resistance in the sentinel practice group versus all general practice was reversed in our study.

There is no consistency in antibiotics showing greater or lesser resistance rates in sentinel practice patients in these studies. However, a common feature of all three is that the differences in resistance rates for sentinel practices compared with overall reported resistance rates are small, and none would justify a change in recommendations for empirical prescribing for UTI in general practice. We conclude that there is good evidence that susceptibility data derived from routine urine samples received by the laboratory provide reliable information for formulating empirical prescribing guidelines for urinary tract infections in domiciliary practice. We suspect that this also applies to specimens from other sites, but this requires further study.