The surveillance, described previously, depends on the voluntary reporting of bloodstream isolates by hospital laboratories in England and Wales (5). The number of laboratories reporting data has grown steadily: by 1998, 208 (91%) of the 229 establishments in England and Wales listed by the Association of Medical Microbiologists were participating. Participation by laboratories in Scotland and Northern Ireland is limited, and their data were excluded from our analysis. Most laboratories used variants of Stokes’ disc method (5) for susceptibility tests in the period reviewed, but a minority used breakpoint tests. Results reported as intermediate were counted as resistant. Quality control was provided by the laboratories’ participation in the National External Quality Assurance Scheme and by comparison to results for the smaller numbers of E. coli isolates from bloodstream infections tested at the Central Public Health Laboratory (6).

Prescribing data for fluoroquinolones, as defined daily doses (7), were estimated for retail pharmacies by using IMS HEALTH’s British Pharmaceutical Index (BPI) and for hospitals by using Medicare Audit’s Hospital Pharmacy Audit (HPA). The BPI records pharmaceutical sales to retail pharmacies and dispensing doctors in the United Kingdom, Channel Islands, and the Isle of Man. Approximately 97% of wholesaler sales to retail and physician outlets and >80% of direct sales by manufacturers are recorded; other sales are estimated from a sample of approximately 600 pharmacies. The number of pharmacies represented in the BPI remained constant during the study period.

The HPA provides information on pharmaceutical consumption by National Health Service hospitals, which account for >95% of hospital care in the United Kingdom. Most hospitals participate: approximately 93% of beds are currently covered. Since 1995, HPA data have been collected monthly from the stock control systems of participating hospitals. Most data are supplied electronically, which minimizes reporting errors. Data include usage of pharmaceuticals among in- and outpatient departments and for private patients in NHS hospitals but not for private patients in designated private hospitals. Before 1995, HPA data were collected from wholesalers, manufacturers, and a panel of hospitals: approximately 90% of indirect sales to hospitals were received from wholesalers and approximately 40% of direct sales from manufacturers. The panel of hospitals covered approximately 80% of beds in 1990 and 84.5% in 1995.

Poisson regression was performed by using the log (total number of isolates with resistance information) as an offset to determine if the proportion of ciprofloxacin-resistant isolates was changing with any type of pattern over time. S-Plus (Mathsoft Inc., Seattle, WA) was used for calculation.

During the 1990s, the Public Health Laboratory Service received nearly 392,551 reports of bacteremia in England and Wales, including 132,311 that indicated E. coli, klebsiellae, Enterobacter spp., and P. mirabilis as the pathogens isolated. These four species groups thus accounted for 32% to 36% of all bacteremia results in each year and for 71% to 72% of those concerning gram-negative bacteria (Table 1). E. coli was the most frequently reported pathogen, causing 22% to 25% of all bacteremias in each year, whereas Klebsiella, Proteus, and Enterobacter spp. were among the 10 most frequent isolates. The number of bacteremia reports rose each year (Table 1), reflecting improved reporting rather than an increased incidence of disease. A fall in the proportion of reports with susceptibility data in 1997 reflected early problems after a switch to electronic reporting and was not exclusive to ciprofloxacin.

Among the reports for E. coli, klebsiellae, Enterobacter spp., and P. mirabilis, 75,168 (56.8%) had susceptibility data for ciprofloxacin, confirming widespread testing. Ciprofloxacin resistance was extremely rare when surveillance began but subsequently increased for all four organisms (Figure 1). The proportion of E. coli isolates reported as resistant rose slowly but steadily, from 0.8% in 1990 to 3.7% in 1999. For Klebsiella spp., the resistance rate rose from 3.5% of reports in 1990 to 9.5% in 1996, before declining to 7.1% by 1999. Enterobacter spp. showed a similar pattern to klebsiellae: the prevalence of resistance rose from 2.1% in 1990 to 10.5% in 1996, then dipped to 7.9% in 1998 before rising to 10.9% in 1999. Only a few P. mirabilis isolates were reported resistant in any year before 1999. Poisson regression showed strong evidence of a trend to increasing resistance for all four organisms and suggested that these increases had a nonlinear component for E. coli, enterobacters, and klebsiellae. If the trends nevertheless were approximated to be linear, the average annual increases in the proportion of resistant isolates were as follows: E. coli, 21.54% (95% confidence intervals [CI] 18.86-24.30); Klebsiella spp., 6.97% (CI 4.41-9.59); Enterobacter spp. 13.97% (CI 10.46-17.58); and P. mirabilis, 21.31% (CI 11.38-32.13).

To assess the distribution of resistance, we counted, for each organism in each year: 1) the number of laboratories reporting resistant isolates, 2) any laboratories contributing >10% of all reports of resistance, and 3) the proportion of reports of resistance from the top three contributors (Table 2). The last two criteria were applied only when >30 resistant isolates of a species were reported in a year, so that a hospital would not appear as a “major contributor” on the basis of three or fewer resistant isolates.

The number of laboratories reporting resistant E. coli rose from 25 in 1990 to 89 in 1999, and no single laboratory ever contributed >10% of all reports of resistance in a year for this species. Laboratories reporting five or more resistant E. coli in years before 1998 mostly served major teaching hospitals, but many district general hospitals reported five or more resistant E. coli isolates in 1998 and 1999. Resistance was more localized and more prevalent in Klebsiella and Enterobacter spp. than in E. coli. The number of laboratories reporting resistant klebsiellae fluctuated from 36 to 57 after 1992, without obvious trend. During a peak in resistance prevalence, from 1995 to 1997, one or two laboratories each contributed >10% of all reports of resistant klebsiellae, and the top three contributors accounted for 32% to 39% of reports of resistance. For Enterobacter spp., laboratories reporting resistance increased from 10 in 1990 to 36 in 1992, then fluctuated with little trend until 1997, before rising to 40 in 1998 and 58 in 1999. In the peak of resistance in 1995 and 1996, two laboratories each accounted for >10% of all reports of resistant enterobacters, and 30% to 32% of reports of resistance came from the top three contributors. Resistance was uncommon in P. mirabilis, and clusters were not evident.

In a further analysis, we identified eight laboratories that frequently reported large numbers of resistant E. coli, Klebsiella spp., and Enterobacter spp. during the entire surveillance period. These were in major metropolitan areas and served teaching hospitals. These laboratories accounted for 7.7%, 11.2%, and 10.3% of reports with ciprofloxacin data for E. coli, Klebsiella, and Enterobacter spp. respectively, but for 18.2%, 30.9%, and 22.4%, respectively, of reports of resistance in these organisms, confirming a major excess of resistance.

The prevalence of ciprofloxacin resistance was examined in relation to patients’ ages for E. coli, since those aged <14 years should not receive fluoroquinolones. Taking the period 1995 through 1999 as a whole, 12 (3.9%) of 305 E. coli with data from patients 1 to 14 years old were reported as ciprofloxacin resistant, compared with 778 (3.2%) of 24,302 E. coli isolates from patients aged >15 years. These data indicated a relative risk of 1.22 (95% CI 0.7-2.1) for the younger patients. Similar calculations were not performed for other species because of the small numbers of source patients aged 1-14 years.

Fluoroquinolone use increased in the earlier years of surveillance, nearly doubling from 1990 to 1993. However, usage has been relatively stable from 1997 onwards, with community use declining slightly (Figure 2). Although most use is still in the community, hospital use has grown steadily in absolute terms and as a proportion, constituting 31.5% of total use in 1999 compared with 18.9% in 1992. Ciprofloxacin was the dominant fluoroquinolone throughout the period (not shown).