An investigation of scattered light integrating collector technology for rapid blood culture sensitivity testing

Abstract

Introduction. Sepsis rates are increasing, with Gram-negative organisms representing a large proportion of bloodstream infections. Rapid antibiotic administration, alongside diagnostic investigations, is required for the effective management of these patients.

Gap statement. Current diagnostics take ~48 h for a final report; therefore, rapid diagnostics are required.

Aim. This study investigated a novel antibiotic sensitivity method, the scattered light integrating collector (SLIC), combined with a rapid identification method using matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) technology to determine if an accurate identification and susceptibility result can be provided within 4 h of a positive blood culture report.

Methodology. A total of 47 blood cultures containing Gram-negative bacteria from 46 patients were processed using the MALDI-TOF Biotyper Sepsityper for identification directly from the blood and the SLIC instrument for susceptibility testing. All organisms were also tested using the current standard workflow used in the host laboratory. Categorical agreement (CA), major errors (MaEs) and very major errors (VMEs) were determined.

Results. SLIC produced susceptibility results with a 71.9% CA, 30.6% MaE and 17.5% VME. The median difference in time to the final result was 44.14 (43 : 05–45 : 15) h earlier compared to the current method.

Conclusion. We conclude that SLIC was unable to consistently provide sufficiently accurate antibiotic susceptibility results compared to the current standard method.

Introduction

Sepsis is a life-threatening condition that affects ~245 000 people in the UK every year, with a mortality of ~20% [1]. Bacteraemia is a common reason for sepsis, and the incidence of confirmed bloodstream infections (BSIs) has increased by 10.8% since 2017 [2]. Gram-negative organisms are frequently isolated from the bloodstream, with Escherichia coli being the most common. Infection with E. coli has increased from 60.4/100 000 population in April 2012/2013 to 67.1 in 2021/2022, indicating that bacteraemia and sepsis are increasing [3].

Rapid administration of empiric antibiotics is required in septic patients to combat an increased risk of mortality and morbidity [4,8]. Targeted therapy using results from culture reduces the risk of toxic side effects from broad-spectrum antibiotics, helps to reduce the impact on the patients’ normal bacterial microbiota and prevents the spread of antimicrobial resistance (AMR) [8,10]. The group of bacteria to which E. coli belongs (Enterobacterales) are currently the major reservoir for AMR BSIs and account for 80.3% of all AMR BSIs. It is therefore important to ensure accurate identification and sensitivity testing of organisms causing sepsis. Such a policy will lead to a reduction in the use of broad-spectrum antibiotics and further retard the development of AMR [2].

Current diagnostic tests for bacteraemia using conventional culture and susceptibility techniques can take 48 h. There is a clear imperative to reduce the time from specimen receipt to result so that antimicrobial therapy can be optimized. More rapid phenotypic [11,17] and genotypic methods [18,20] have been proposed.

In this study, we set out to evaluate the utility of the scattered light integrating collector (SLIC) as a rapid automated sensitivity testing device for BSIs in a clinical laboratory setting. The SLIC instrument uses static laser light scattering to determine the phenotypic susceptibility of a cultured bacterium to antibiotics within 4 h. A laser is directed through a sample that reflects, refracts and diffuses, depending on if the sample has scattered or absorbed the photons. If there is growth of the organism, there will be increased scattering of photons. A photodiode measures the scattered photons and reads the signal as the change in an electric current, which is recorded over time. A suspension of bacteria scatters the light based on the number and morphology of the cells within the suspension [21].

Here, we examine the performance of SLIC in determining antimicrobial susceptibility profiles of Enterobacterales from positive blood cultures compared to conventional methods in a clinical laboratory setting. Time to the availability of the result, as well as any hypothetical antimicrobial management changes, was also examined.

Methods

This comparative diagnostic study was carried out in Lancashire Teaching Hospitals NHS Foundation Trust, a secondary care hospital in the North-West of England with ~900 beds across two sites, between 9 June 2021 and 17 December 2021. Forty-seven positive blood culture samples from 46 patients, which were flagged positive by the BacT/ALERT system and contained Gram-negative bacilli on microscopy on the study days, were included and processed using the current laboratory method as well as the study method (SLIC). Exclusions included any non-Enterobacterales isolates and those that had their SLIC testing interrupted (Fig. 1).

Fig. 1. Breakdown of samples for SLIC and clinical analysis.

Current laboratory method

Blood cultures were incubated in the BacT/ALERT® 3D Microbial Detection System (bioMérieux). Positive cultures were processed as presented in the workflow in Fig. 2, with Gram stain, direct antimicrobial susceptibility via disc diffusion [European Committee on Antimicrobial Testing (EUCAST)] and sub-culture for identification the following day using mass spectrometry (MALDI Biotyper®, Bruker). Definitive susceptibility results were obtained using a VITEK®2 (bioMérieux) instrument.

Fig. 2. The current laboratory positive blood culture workflow for Lancashire Teaching Hospitals NHS Foundation Trust using the direct sensitivities, MALDI-TOF MS for identification and the VITEK®2 (with pivmecillinam and trimethoprim using EUCAST disc diffusion) for sensitivity testing as well as the study method workflow.

SLIC

Organisms from positive blood culture bottles were identified directly from liquid culture using a matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) Sepsityper, according to the manufacturer’s recommendations and as previously described [22]. Liquid media from positive blood culture bottles were centrifuged at 1500 g for 4 min to remove inhibiting substances, prior to the supernatant being diluted 1 : 10 in sterile PBS. This sample was then added to 880 µl of media to make a 1 : 500 dilution in semi-micro-optical cuvettes, with the final antibiotic concentrations (EUCAST resistant breakpoint MIC), as detailed in Table 1 [23]. Positive (patient sample and media only) and negative (uninoculated media) controls were also included. Sensitivity testing was carried out on SLIC for 3 h, at which point output data were produced.

Table 1. Final concentrations of antibiotics in the SLIC analysis.

Antibiotic	Final cuvette solution concentration (µg ml−1)
Amoxicillin	8
Pivmecillinam	8
Co-amoxiclav	8
Piperacillin–tazobactam	8
Cefuroxime	8
Ceftazidime	4
Ceftriaxone	2
Cefpodoxime	1
Cefoxitin	8
Ertapenem	0.5
Meropenem	8
Gentamicin	2
Ciprofloxacin	0.5
Trimethoprim	4

Data analysis

The data produced by the SLIC instrument were run through a specifically designed macro. This determined the sensitivity of an organism to an antibiotic when there was a 50% reduction in its growth compared to the positive control. The identification results of the study method were compared with the current laboratory method, and the percentage identification match was recorded. The sensitivity testing of the study method using SLIC was compared with the chosen ‘gold standard’ (VITEK®2 for all antimicrobials except pivmecillinam and trimethoprim, which were tested by EUCAST disc diffusion). The SLIC susceptibility result was assessed for categorical agreement (CA), in which the percentage of isolates providing concordant results from SLIC and the current method was calculated. Major errors (MaEs) were defined as a resistant result by SLIC and a sensitive method by the current method. Very major errors (VMEs) were defined as a sensitive result by SLIC and a resistant result by the current method [24,28].

Clinical results

A clinical microbiologist retrospectively accessed the prescribing data for each patient and assessed if the patient was on appropriate antibiotics at the time of the positive blood culture. The time taken to the final identification and susceptibility report was calculated from the time the Gram stain was reported to the clinical teams. These results were assessed for normality using Kolmogorov–Smirnov and Shapiro–Wilk tests in SPSS® and were found to be not normally distributed. A Wilcoxon-Signed Rank-Sum test was therefore used to assess for a statistically significant difference in time taken to the final report between the existing method and the SLIC study method. The null hypothesis was that the difference in time taken to the final report was zero.

Results

Identification

Forty-seven samples from 46 patients were processed using the current laboratory method and the study SLIC method. Of these, E. coli was isolated the most often (57.4%), followed by Klebsiella pneumoniae (8.5%) and Proteus mirabilis (8.5%), with other Enterobacterales making up the remainder. There was a 100% agreement in identification between the study method and the current method.

Antibiotic susceptibility testing

Overall, the CA between the current laboratory method and SLIC was 71.9%. There was no antibiotic that produced 100% concordance, although gentamicin and amoxicillin performed at 93.6 and 90%, respectively (Table 2). Piperacillin–tazobactam performed the worst, providing a concordance of 17.0%, whilst the remainder of the antibiotics tested were above 50%. The median number of MaE was 8 (Inter Quartile Range [IQR]: 5–13), with all antibiotics having at least one MaE reported. Amoxicillin had the fewest, with 2, and piperacillin–tazobactam had the most, at 38. The median number of VME was 1 (IQR: 0–2), and no VMEs were reported for four antibiotics (ceftazidime, ertapenem, meropenem and gentamicin), with six for cefoxitin as the highest number. In total, out of 604 tests, there were 150 MaEs (30.6%) and 20 VMEs (17.5%).

Table 2. Data and categorical agreement from the SLIC method compared with current laboratory methods for each antibiotic.

Antibiotic*	Categorical agreement	Major errors (MaEs)†	Major error rate (%)	Very major errors (VMEs)‡	Very major error rate (%)
Amoxicillin	90.0% (36/40)	2	12.5	2	8.3
Pivmecillinam	72.3% (34/47)	10	25.0	3	42.9
Co-amoxiclav	53.2% (25/47)	21	77.8	1	5.0
Piperacillin–tazobactam	17.0% (8/47)	38	90.5	1	20.0
Cefuroxime	80.9% (38/47)	7	21.9	2	13.3
Ceftazidime	76.6% (36/47)	11	25.0	0	0.0
Cefpodoxime	70.2% (33/47)	13	33.3	1	12.5
Cefoxitin	78.7% (37/47)	4	11.1	6	54.5
Ertapenem	57.4% (27/47)	20	42.6	0	0.0
Meropenem	89.4% (42/47)	5	10.6	0	0.0
Gentamicin	93.6% (44/47)	3	6.8	0	0.0
Ciprofloxacin	78.7% (37/47)	8	19.5	2	33.3
Trimethoprim	78.7% (37/47)	8	22.9	2	16.7
Total	71.9% (434/604)	150	30.6	20	17.5

1*Ceftriaxone was the 14th antibiotic tested on SLIC; however, this is not tested on the routine laboratory panel, so it was excluded from this analysis. Amoxicillin has fewer total tests compared to the other antibiotics, as some organisms are not tested for amoxicillin due to intrinsic resistance (e.g., Citrobacter species,and Enterobacter species).

2†MaE rates are calculated using the total number of susceptible organisms.

3‡VMEs are calculated using the total number of resistant organisms.

Timings

The median time taken for patients to be placed on effective antibiotics was 0 min (0–232 min) due to a large number of patients in the cohort (62) already being treated with effective antibiotics when the blood culture was reported as positive.

Using the current method in the laboratory, the median time (IQR) for the final identification and sensitivity result to be available was 47 h 48 min (46 : 38–48 : 45). In comparison, the study method found that, for the 47 samples tested for 3 h, the median time (IQR) was 3 h 35 min (3 : 30–3 : 40). When compared to the median time for the current method, using SLIC would allow for the final identification and sensitivity report to be available 44 : 14 h earlier (43 : 05–45 : 15) (P=1.4e-14), which therefore means that the null hypothesis that there is no difference in time taken to the final report between the two methods was rejected.

Discussion

Overall, the performance of SLIC was not comparable to that of the current workflow, with a total CA for all the tested antibiotics of 71.9%. The study method did, however, produce a result within the 4-h objective and nearly 2 days earlier than the current standard method. This CA result would mean that this method does not meet the current criteria necessary for clinical implementation. This is because the CA for the introduction of a new susceptibility method should be 90% and above for acceptance [28,30]. The acceptable rates for MaE and VME vary within the literature: the VME rate depends on the number of resistant isolates tested, although the generally accepted values are <1.5% for VME and <3% for MaE [27,30].

The first and second most frequently isolated bacterial species in the study correspond to the current first and second most common Enterobacterales isolated from blood cultures in England (E. coli and K. pneumoniae), with Proteus species the third most common [2]. The 100% concordance with the Sepsityper MALDI method is in agreement with literature reports, which show >90% concordance with organism identification using blood directly from the positive blood culture bottle [31,34]. Identification of the pathogen can be useful for the management of patients, as some organisms are known to have specific resistance patterns. Given, however, that the most commonly isolated organisms were E. coli, K. pneumoniae and P. mirabilis, none of which are known to have intrinsic mechanisms conferring clinically significant resistance; having this identification result for these organisms does not help with any expected sensitivity results.

The SLIC instrument did not perform as well as expected compared to a previous study carried out on positive blood cultures by the Hammond Research Group (University of St Andrews), where an average CA of 95.5% and the rate of errors of 4.1% MaE and 7.1% VME for a 2-h analysis were demonstrated [35]. The previous study was conducted for 2 h, not 3, but, in this study, the CA was 71.9% and the MaE rate was 30.6%. The VME rate was 17.5% when the SLIC data were collected after 3 h. This reduction in CA compared to the previous study on SLIC is unexpected, although the study methods were not the same. The SLIC instrument provided for this study was a bespoke 16-well model, and an increased number and range of antibiotics were used. Both methods used only Enterobacterales, with the majority being E. coli in both studies. The differences in instrument, timing and antibiotic panel and the fact that this study tested more antibiotics may explain the difference in results produced. In this study, the identification of piperacillin–tazobactam resistance was responsible for the majority of errors; this was not tested in the previous study, which may explain some of the differences in results. In addition, although extensive validation studies had been carried out on the 6-well instrument, the 16-well instrument was the first of its kind and was only validated in its capacity to detect the growth of E. coli [American Type Culture Collection (ATCC) strain 25922] and Staphylococcus aureus (ATCC stain 25923) grown in brain heart infusions and Mueller Hinton broth, which may explain the disparity between the two instruments.

Of the antibiotics tested, the poorest performer was piperacillin–tazobactam, and the best was gentamicin. These observations are similar to a study using Alfred 60AST (Alifax), which found the most discrepant results for piperacillin–tazobactam and managed to correct some of these with modification of the drug formulation by the manufacturer [36]. The resulting variation dependent on the method is said to be a known problem for piperacillin–tazobactam, which may explain the poor performance on SLIC [37].

There were variable CA results across the different antibiotics tested. One potential reason might be attributable to the fact that true MIC results can be one dilution on either side of the reported result. This could change whether the result is reported as resistant or susceptible, which could affect a comparative study between the two methods. There could also be variation due to the antibiotic stability when the dilutions were being prepared, frozen and then thawed, which may vary between antibiotics. The data were reviewed to examine whether bacteriostatic or bacteriocidal categorization had an influence on the results, but this does not appear to be the case: 12 of the antibiotics were bacteriocidal, and trimethoprim was the only bacteriostatic drug.

This SLIC instrument did not perform favourably in comparison with the standard method, which, unfortunately, is not what has been seen in other rapid susceptibility testing platform analyses. Platforms such as QMAC-dRAST™ (based on microfluidics), Accelerate Pheno®, Specific Reveal™ and Alfred 60AST (Alifax) have shown better performance with CA of 90.6–97.9%, VME of 0.0–3.3%, MaE of 0.0–2.5% and ME of 0.7–7.4% within 7 h [12,14, 17, 36, 38,54].

In this study, the median time to effective antibiotics was zero, indicating that the patients were already on effective antibiotics in the majority of cases. Empiric antibiotics for sepsis are based on local epidemiology and are designed to cover the most common organisms in a specific clinical context: locally, this does not include a high number of carbapenem-resistant Enterobacterales. This reasoning does not consider if the antibiotics were appropriate or unnecessarily broad spectrum. This was also seen in other studies, where patients were already on effective antibiotics 75–94% of the time [42,55], although this is contrary to other studies that report lower numbers of patients already on appropriate antibiotic therapy. This may be related to the antibiotic resistance rates and choice of empiric treatment in the local study area [6,42, 56,58]. A high percentage of patients undergoing treatment on effective antibiotics does not necessarily support the case for rapid susceptibility tests—the patients are being treated for the infecting organism already. The availability of rapid susceptibility results, however, can allow for the use of narrow-spectrum antibiotics. This can potentially reduce side effects such as Clostridium difficile and drug toxicity, and improve time to oral switch, which can facilitate hospital discharge and reduce length of stay. In addition, having information to allow for a de-escalation of antibiotics helps with antimicrobial stewardship and development of resistance.

Rapid identification and sensitivity results can potentially benefit a patient clinically; however, a limiting factor that needs to be considered is whether the rapid results are going to be acted upon clinically once the testing is carried out. Many laboratories, including the one hosting this study, have specific opening hours of 8 am–6 pm on weekdays and 8–1 pm on weekends, and positive blood cultures are not processed outside of these hours. This may negate the benefit of the rapid positive result if it becomes available out of hours. This is a factor that would need to be taken into consideration if introducing a rapid method into the department.

There are several limitations to this study, including the fact that, although the VITEK®2 was noted as the ‘gold standard’ for the majority of the antibiotics tested in this study, it is not free from error itself. There may have been cases where the VITEK®2 result was incorrect and the SLIC result was correct because the VITEK®2 was not 100% accurate. VITEK®2 has previously been shown to have a minimum inhibitory concentration agreement (and therefore CA) of 84.2–95.6 %, a VME of 0.4% and MaE of 0.5% when compared to broth microdilution [59]. This was not investigated within this study, and the VITEK®2 results were used as the defined laboratory standard. Initially, discrepancies were going to be investigated further, but unfortunately, due to the high number, this was logistically not possible. In addition, there were only 47 samples tested for 3 h; additional results may help to determine the performance of this method. The study also did not look at mixed cultures, which can often happen within the laboratory or Pseudomonas spp. isolates: both would need to be assessed in the future. This study focussed on Enterobacterales—future studies should include all Gram-negative and Gram-positive organisms. A larger sample size, with a wider array of organisms, would allow the results to be generalized to all positive blood cultures; these results can only apply to Enterobacterales. Pre-analytical parameters were not assessed, including the time taken for the blood culture to get to the laboratory (this is being addressed in the current Standards for Microbiological Investigation). There was also an absence of carbapenem-resistant organisms and a low incidence of gentamicin resistance, which means that these antibiotics have not been tested thoroughly; the results for these antibiotics may therefore be skewed. Further study is required to provide additional information on the performance of SLIC, and improvements to the method can be made. The antibiotic solutions were created by diluting powders along with a freezing and thawing cycle, which could introduce inaccuracies. In order to overcome this, alternative methods, such as impregnated discs in the cuvettes or antibiotic-impregnated discs, could be used in the future. The samples were incubated for 3 h to gain the sensitivity results: an alternative option would be to incubate for longer to assess if this intervention changes the overall results.

Conclusion

The system under investigation during this study (SLIC) has not performed sufficiently well to meet the standards necessary for clinical implementation at the present time without further investigation. However, the results do suggest that rapid susceptibility results could provide a final report significantly quicker than the current standard method, and this could potentially lead to more targeted antibiotic treatment, although this would need to be studied further.

Abbreviations

AMR

antimicrobial resistance

ATCC

American Type Culture Collection

BSIs

bloodstream infections

CA

categorical agreement

EUCAST

European Committee on Antimicrobial Testing

HCRW

Health Research Agency and Health and Care Research Wales

IQR

inter quartile range

MaEs

major errors

MALDI-TOF MS

matrix assisted laser desorption/ionisation -time of flight mass spectrometry

MIC

minimum inhibitory concentration

PBS

phosphate buffered saline

SLIC

scattered light integrating collector

VMEs

very major errors