Clinical impact of a commercially available multiplex PCR system for rapid detection of pathogens in patients with presumed sepsis

Early adequate antibiotic treatment improves the outcome of patients with sepsis [1‐5]. Even if broad spectrum antibiotics are used empirically, adjustments of antimicrobial therapy may be necessary. Generally adjustments are based on the results of positive blood or other cultures that are available after 8 to 48 hours [6].

Additionally, the likelihood of a positive result in conventional culture methods can be reduced by concomitant or prior antibiotic treatment. Amplification of bacterial and fungal nuclear acids directly from blood specimen is a newly established detection method for pathogens. Tests based on this method may improve clinical care by shortening the time to a positive result and by being more independent of antibiotic pre-treatment. The impact of these methods on therapeutic decisions and outcome has not yet been studied.

We compared the results of a standard BC system with those of a commercially available polymerase chain reaction (PCR) based system in routine clinical management.

Primary goal of the study was to determine if results from the additional PCR based system led to different therapeutic decisions than the results from BC alone.

Secondary goal was to assess the concordance of both methods and time saving effects with the use of a PCR-based test.

A commercially available molecular based test system (LightCycler^® SeptiFast^® (SF) Test; Roche Diagnostics, Mannheim, Germany) was offered as an add-on diagnostic tool free of charge supplemental to standard BCs taken in patients with presumed sepsis at the Regensburg university medical centre during an eight months period (July 2006 – March 2007). The test was made available to all departments treating patients with sepsis by the microbiology department. The decision to use the additional test was solely made by the treating clinician.

The SeptiFast^® Test is able to detect 25 different pathogens directly from blood by real-time multiplex PCR (Table 1) [7]. The limits of detection were 100 CFU/ml for coagulase-negative staphylococci, C. glabrata and Streptococcus spp. and 30 CFU/ml for all other pathogens listed in the table [8].

Both the sample for the BC and the sample for the SF were taken through the same blood drawing device. Both aerobic and anaerobic BC bottles (Bactec 9240 BC system, Becton Dickinson, Heidelberg, Germany) were inoculated directly with 10 ml blood each and delivered to the microbiology department together with the aliquot for analysis with the SF. Blood cultures were incubated for 7 days. The SF-test was provided free of charge by the manufacturer.

A retrospective analysis of the test results and clinical data was performed. Clinical data were extracted by chart review (underlying illness, antibiotic pre-treatment, epidemiological data such as age, sex, length of stay, hospital mortality, immunosuppression, SIRS criteria (Temperature ≤ 36°C or ≥ 38°C, heart rate ≥ 90 bpm, respiratory rate ≥ 20 breaths/min or paCO₂ < 32 mmHg, white blood cell count ≥ 12,000 or ≤ 4,000 cells/mm³)) and by the data provided for the test application. Immunosuppression was defined as being post organ-transplant, receiving chemotherapy for malignant disease or receiving high dose prednisolone (>20 mg/day). Those with chronic disease predisposing for infections such as chronic liver disease or alcoholism were not counted as such. Antibiotic pre-treatment was defined as having received any antibiotic therapy within 24 hours before specimen collection. Multiple samples per patient were allowed, if the times of blood collection were at least 48 hours apart.

To determine the impact of the results on the patients' therapy, data were independently reviewed by two infectious disease specialists. They analyzed the current antibiotic therapy and possible adaptations necessary for optimal treatment according to current standard of care. In case of both SF and BC yielding positive results, the possible time saving effect of the SF-test was determined.

Treatment adjustments were classified into two different categories: 1) Treatment adjustment not necessary (either due to coverage by the empirical therapy or to lacking clinical relevance); 2) Treatment adjustment necessary due to lacking coverage of empirical therapy.

In order to examine possible time saving effects of the molecular based analysis, time to results of BCs was compared to the time of SF results. At our institution BC results are reported to the treating physician by telephone call from 8 am to 7 pm Monday-Friday and 9 am – 4 pm on weekends and holidays. A complete report including antibiotic resistance testing results is faxed and sent by mail. Due to the experimental status of the SF-test, the results were not available on a daily basis. Therefore a virtual once and twice daily analysis Monday-Friday was assumed in order to achieve comparable time frames. For the once daily simulation results were assumed to be available by 2 pm if received by 8 am in concordance with the turnaround times reported for the test. In the twice daily setting results were assumed to be available by 2 pm if SF samples were received by 8 am, if received by 1 pm results were assumed to be available at 7 pm. We then compared the time from drawing the blood samples to the time a first positive BC result was reported and the virtual SF result time.

Data collection and statistical analysis were performed using Microsoft Office Access^® (Version 2007, Microsoft Corporation, Redmond, USA) and SPSS^® (SPSS Inc, Version 15, Germany).

Due to the retrospective and non-interventional design of the analysis approval of the local ethics committee was not required according to the guidelines of our hospital.

In 39/101 samples from patients with suspected sepsis pathogens were identified with either the use of BCs or a PCR-based test for the identification of bacterial and fungal pathogens in whole blood (SF-test).

Concordance for positive and negative results between BC and SF (77%) was comparable to other studies [9, 10].

One important question is if these pathogens additionally detected by either SF or BC are clinically relevant or if there are merely "innocent bystanders", i.e. the important discrimination between contamination, colonisation or true infection. In our patient cohort most pathogens detected only by SF were also present in other specimens of the same patient suggesting a clinical relevance. These results are in concordance with other studies where 12 of 14 respectively 31 of 42 pathogens additionally detected by PCR were present in other samples of the same patient [9, 10].

CNS were rarely detected by the SF-test in our collective, as the SF-test tries to omit positive results due to contamination. This is in accordance with one other study, presenting 102 patients without signs of infection with 19.6% positive BC and no CNS by SF [7]. In our study in all patients with CNS identified by BC the antibiotic treatment was not adjusted, alluding to the considered insignificance by the treating clinicians. Whether at least one of them was clinically relevant could not be judged retrospectively.

Our study showed that the use of SF in addition to BC led to the identification of more pathogens. If this was influenced by concomitant antibiotic therapy could not be addressed in our study, as 82% of the patients were already on treatment at specimen collection.

Our assumed twice daily analysis for the SF-test would have led to a median time to final results of 18 hours. Compared to usual turnaround times for BC results, this yields a clear time advantage for the SF-test. Immediate processing of specimens could further shorten the time to final results to 6.5 h [10]. If this approach is feasible in a routine hospital setting has yet to be examined.

Additional pathogens were identified in 5 cases and in 3 samples the rapid SF-test would have led to earlier adjustment of antibiotic therapy. In total the use of the test would have resulted in therapeutic consequences in 8 (8%) patients.

On the other side, 9 pathogens were not detected by SF. One of them is not included in the PCR portfolio, thus no conclusions can be drawn regarding sensitivity. Alarming is the failure of detection of S. pneumoniae, even if this could be attributed to a software problem that was solved in newer software versions. With regard to the missed polymicrobial infections it is unclear why these were not detected by SF. A possible explanation might be the competitive growth in vivo that does not allow both pathogens to reach a level resulting in a positive SF test.

All detection failures and the necessity of antibiotic resistance testing underline the importance of using the SF test additionally to standard BC.

If therapeutic adjustments initiated by the SF-test have an impact on the survival of patients with sepsis could not be determined in this study. Studies indicate that patients do not have a benefit from changes in antibiotic therapy if the initial empiric therapy did not cover the causative pathogen [4]. As those studies are based on the results obtained with current culture systems an average delay of 1–2 days has to be assumed [6]. The effect of shortening this time interval with the use of a PCR test has yet to be examined [11].

Invasive fungal infections have an especially high mortality [4, 12, 13]. Studies have demonstrated that adverse patient outcomes are related to late or inadequate antibiotic treatment. As the SF-test was superior in detecting fungal pathogens and delivers results more rapidly, the use of SF may improve patient outcomes.

The cost benefit of the SF may differ in different clinical situations or patient populations. Special equipment and specially trained personnel are necessary to perform the test. The cost of individual tests will be influenced by the intended turnaround time as well as by the availability outside normal working hours. Cost and cost-effectiveness analyses will have to be done for different scenarios.

Our study had several limitations. Only 80% of the patients presented with two or more SIRS criteria at the time of the specimen collections. But as the rate of immunosuppressed patients was high, a high index of suspicion was present in the remaining 20%. Additionally, it was a non-randomized study, where the decision to use the additional test was solely taken by the treating clinician. Thus a selection bias has to be assumed that may influence the rate and types of pathogens found in our samples. Further, it is a single centre study with a small study size, thus possibly limiting the generalisability of the results. Nevertheless the design of our study may mirror clinical reality more exact than randomized controlled trials. The patients' characteristics in our study are well comparable to those of patients from sepsis studies or cohorts, including one at our institution [14].

Our study shows that the use of a rapid molecular diagnostic test for the identification of bacterial and fungal pathogens may lead to a higher rate of early adequate antibiotic therapy in approximately 8% of patients with suspected sepsis. Due to the small sample size and the design of our study cost effectiveness analyses were not feasible. They are warranted in further studies.