METHODS AND MATERIALS

Media and reagents

CLED agar (CM301; Oxoid, Basingstoke, UK) and CUTI medium (CM949; Oxoid) were produced according to the manufacturer's instructions. All plates were produced at one site and subjected to full quality control procedures before distribution. All plates were used within 14 days of preparation on the basis of pre-trial shelf life testing, which showed no decline in recovery rates or colony size over this period.

Indole reagent (p-dimethylaminocinnamaldehyde; Sigma, Poole, Dorset, UK) was produced as a 1% wt/vol solution in 10% hydrochloric acid. This was quality controlled using positive control E coli (NCTC 9001) and negative control Enterobacter cloacae (NCTC 11936).

The following reagents were also used: filter paper strips (BTR 1; Mast, Bootle, Merseyside, UK), API 20 E and 20NE (Bio-Merieux, Basingstoke, UK), BBL Crystal (Becton-Dickinson, Oxford, UK), Microbank^TM Cryo beads (PL. 160; Prolab, Wirral, UK), cephalexin 30 μg disks (CL30; Oxoid), and the OBIS PYR test (ID580M; Oxoid). These were all used according to manufacturers' instructions.

Samples

Samples were selected on the basis of a white blood cell count (WBC) of > 100 × 10⁶/litre. A total of 1466 routine samples received during a two week period in January and February 2000, from both hospital and general practice, were included in our study.

Media inoculation and incubation

Using a 1 μl loop, all 1466 samples containing > 100 × 10⁶/litre WBC were inoculated on to ¼ plate CLED and CUTI medium. In addition, all the samples were inoculated on to square CLED and CUTI agar plates using the filter paper strip method. All plates were incubated in air at 37°C ± 1°C for 18–24 hours.

Plate reading

The staff of the laboratories previously inexperienced with the use of CUTI medium were given two weeks of training before the study to ensure familiarity with colony appearances. CLED and CUTI media plates were read independently and by different members of staff.

Data on bacterial growth and significant isolates were recorded at the time of reading and discrepancies verified daily by a separate senior member of staff. Bacterial growth was recorded semiquantitatively on the basis of colony counting. For loop inoculated plates, one colony equated to 10⁶ cfu/litre, increasing to > 10⁸ cfu/litre for more than 100 colonies. For filter paper strip cultures, between one and five colonies equated to 10⁷ cfu/litre, increasing to 10⁸ cfu/litre for 20 colonies.

For the purpose of our study, samples containing no more than two organism types were considered significant. Samples containing three or more organism types were classified as mixed growths and no further investigations were performed.

Isolates from CLED agar were recorded according to generic classification with bench tests performed as applicable—that is, the staphylococcal latex test for coagulase clumping and/or protein A, oxidase, catalase test, and novobiocin sensitivity. Isolates from CUTI agar were recorded according to colour as described in the manufacturers' instructions. As recommended, spot indole using DMACA reagent (Sigma) was performed on all pink colonies.

Significant Gram negative bacilli isolates from the loop inoculated CLED and CUTI medium were saved on to Microbank Cryo beads (Prolab) and stored at −70°C for subsequent identification.

Identification

All Gram negative isolates were identified using two commercial systems, Bio-Merieux API (20E and 20NE) and Becton-Dickinson Crystal (E/NF). Isolates giving discrepant results were submitted to the Identification Services, Central Public Health Laboratory, Colindale, London, for identification by additional methods.

Supplementary tests

All organisms identified as Citrobacter spp and a random 4% selection of E coli were further tested for cephalexin susceptibility and l-pyrrolidonyl-β-napthylamide (PYR) activity. All organisms identified presumptively as Proteus spp were tested for indole production.

Statistical methods

Data were collated centrally and entered into an Excel spreadsheet. Differences between laboratories were compared using the χ² test and differences between methods using McNemar's test.

RESULTS

Method/media comparisons

One thousand, four hundred and sixty six urine samples were examined from the four laboratories. Statistical analysis indicated no significant difference between the results submitted from the four sites and therefore the amalgamated results are presented here.

Tables 1– 4 and figs 1–4 summarise the isolation results and the detection of mixed cultures.

The comparison of CLED with CUTI for the detection of mixed cultures showed a highly significant improvement in detection by both the loop (p < 0.001) and filter paper strip (p = 0.030) methods using CUTI medium (tables 1,2; figs 1, 2). It can be seen that for both methods CUTI gave significantly fewer samples with no growth (p = 0.014 and p = 0.013), indicating that CUTI supported the growth of more urinary isolates than CLED.

A direct comparison of the loop and filter paper strip methods (table 4; fig 4) confirmed the superiority of the loop method in detecting mixed cultures on CUTI medium (p < 0.001), but there was no significant difference between the inoculation methods on CLED medium (p = 0.14; table 3; fig 3)

Overall, the CUTI loop inoculated method identified more mixed growths than the filter paper strip methods (p < 0.001) and hence fewer significant isolates (p = 0.012). Fewer plates were recorded as no growth by this method (p = 0.077; table 4; fig 4).

Identification

Using the CUTI loop inoculated results, of the 1466 samples tested, 973 samples (66.4%) gave significant isolates (≤ 2 isolates/sample), resulting in 676 saved Gram negative isolates, with the remaining isolates consisting of a variety of organisms including staphylococci, streptococci, yeasts, and lactobacilli.

Identification of the Gram negative isolates using Bio-Merieux API 20E/NE and Becton and Dickinson Crystal E/NF gave > 99% (672 of 676) agreement and therefore the amalgamated results are presented here (see fig 5 for a summary identification of isolates).

Using the colour criteria provided by Oxoid, the correlation of presumptive identification from CUTI medium with confirmed identification using API/BBL is summarised in table 5.

It can be seen that the CUTI medium in combination with the bench indole test was 99% specific for E coli but this was reduced to 97% if the indole test was omitted because 11 of 15 Citrobacter spp gave false presumptive E coli identification (pink colonies). Although the indole test eliminated all Citrobacter freundii isolates (seven of the 11 Citrobacter spp), it was not sufficient to eliminate the misidentification of Citrobacter koseri. To perform indole tests to eliminate the misidentification of seven of 15 Citrobacter spp would require the testing of all pink colonies (74.3% of all isolates). It might be considered that this specificity gain is not sufficient to warrant the large number of indole tests that would need to be performed.

It was anticipated that the misidentification of Citrobacter spp as E coli would affect the apparent prevalence of cephalosporin resistance as a result of chromosomal β-lactamases, which would affect surveillance data. On the evidence of previous studies, it was expected that 80% of C freundii would be resistant to cephalexin,¹³ and therefore cephalexin resistance could be used as an alternative marker of a need for further identification. However, supplementary testing of the pink citrobacter isolates (one isolate of C koseri and 10 isolates of C freundii) demonstrated that approximately 50% were resistant (six isolates of C freundii) to cephalexin. These results demonstrate that reliance on cephalexin resistance as a marker in the absence of an indole test would still misidentify half of the citrobacter isolates as E coli.

If cephalexin sensitivity is used to increase the specificity of E coli identification of pink colonies, a PYR test would be required to differentiate the cephalexin resistant isolates as E coli or C freundii. Further testing of the E coli and citrobacter isolates from the study confirmed that all the E coli isolates tested were PYR negative and citrobacter isolates were PYR positive.

Using the PYR test and cephalexin susceptibility, the sensitivity and specificity achieved for E coli identification were 97.2% (487 of 501) and 98.2% (487 of 496), respectively.

The chromogenic medium did not differentiate between isolates presenting as blue/purple colonies. These included species of klebsiella, enterobacter, and serratia and further identification tests would be required.

One hundred per cent sensitivity and specificity for Pseudomonas spp were achieved using the criteria of isolates being colourless and oxidase positive. Ninety four per cent of the brown colonies were species of proteus/morganella. Negative indole tests accurately distinguished Proteus mirabilis, which comprised 2% of the isolates.

Overall, 86.2% of isolates could be identified using the chromogenic medium, supplemented by cephalexin susceptibility and oxidase tests. Table 6 summarises these data. The remaining 13.8% of isolates would require further identification.

DISCUSSION

Semiquantitative plating of urine samples on to CLED agar has been an integral part of the microbiological investigation of UTI for many years. Previous comparisons of methods comparing loop inoculation with filter paper strips have indicated that both are suitable as a means of identifying infection. Our study has confirmed that for the detection of significant isolates there was no significant difference between loop and filter paper strip inoculation methods when using CLED agar.

Conversely for CUTI medium, there were significant differences in the detection of mixtures. A higher proportion of significant isolates by the filter paper strip method corresponded to a higher proportion of mixed cultures by the loop method, confirming the superiority of the loop method to detect mixed cultures.

If loop inoculation is used, the impact of changing from CLED to CUTI is to decrease the number of significant isolates from 1074 to 1034 (p = 0.011). This decrease of 40 is the result of an increase in the detection of mixed cultures. If the filter paper strip method is used the impact of changing from CLED to CUTI is to decrease the number of significant isolates from 1073 to 1068, but this result was not significant.

The increased number of mixed cultures detected on CUTI medium was largely accounted for by the presence of Gram positive organisms, especially enterococci. Because the main objective of the research was to validate the performance of CUTI medium for Gram negative identification this was not formally assessed as part of the trial.

“Our study has confirmed that for the detection of significant isolates there was no significant difference between loop and filter paper strip inoculation methods when using cystine lactose electrolyte deficient agar”

Commercial identification systems (Bio-Merieux API 20 E and Becton-Dickinson BBL Crystal) performed equally well for the identification of Gram negative isolates.

Using the supplementary bench tests, 86% of Gram negative isolates could be identified on CUTI medium. The identification of E coli, Pseudomonas spp, and P mirabilis could be confirmed by simple spot tests.

Citrobacter spp were commonly misidentified as E coli and, as described by Chagla et al,¹⁴ the bench indole test failed to differentiate the Citrobacter spp from E coli. Improved differentiation between Citrobacter spp and E coli has been demonstrated previously by Chagla et al using a PYR hydrolysis test.¹⁴ Supplementary testing of the saved isolates using the Oxoid PYR kit confirmed that all of the Citrobacter spp isolated were PYR positive compared with the negative PYR results obtained from the E coli isolates. However, to undertake a PYR test on all pink colonies would be much more expensive than performing an indole test in a routine laboratory. It had been suggested that by screening for cephalexin susceptibility and performing a PYR test on resistant isolates, the misidentification of Citrobacter spp as E coli would be resolved, but this was found to be only partially correct. Of the 11 pink Citrobacter spp isolated, six were found to be cephalexin resistant and could therefore be correctly identified with a supplementary PYR test. The remaining five cephalexin sensitive, pink Citrobacter spp would be falsely identified as E coli, which resulted in an E coli specificity of 98.2%.

Further work to develop a reliable and inexpensive identification system for the remaining 13.2% of Gram-negative isolates is required.

The cost implications of using pre-poured plates of CUTI medium as a direct replacement for CLED medium into a routine laboratory testing 50 000 samples each year would be an additional £3750. However, this needs to be assessed against the cost of undertaking less susceptibility testing (more mixed growths) and the added value of gathering reliable antibiotic surveillance data.

In summary, CUTI medium although more expensive than CLED, improves the detection of mixed cultures and when supplemented with simple bench tests provides an identification system that is capable of speciating 86.8% of Gram negative isolates, facilitating antimicrobial resistance surveillance for E coli and P mirabilis.