A Single Negative Result for van Quantitative PCR on Enrichment Broth Can Replace Five Rectal Swab Cultures in Screening for Vancomycin-Resistant Enterococci

ABSTRACT

The increased incidence of infections by vancomycin-resistant Enterococcus (VRE) causes an accumulation of patients who are either colonized with VRE or flagged as potentially colonized with VRE. Since such patients require precautionary isolation upon admission to a hospital, rapid methods to establish VRE colonization status would improve patient care and optimize hospital operation. We evaluated van quantitative PCR (qPCR) on one enrichment broth as a VRE-screening approach. We obtained 255 sets of five rectal specimens from 243 patients. The specimens were cultured using an amoxicillin-containing enrichment broth. Subsequently, a chromogenic agar was incubated and suspect colonies were inoculated on a blood agar plate and characterized by matrix-assisted laser desorption ionization–time of flight (MALDI-TOF), followed by a vancomycin Etest in cases in which Enterococcus spp. were detected. The culturing results were compared with the outcome of van qPCR on all enrichment broths of the first rectal swab. The van qPCR was positive for 43% of the sample sets (vanA, n = 5; vanB, n = 101; vanA and vanB, n = 3). Based on culture data, 20 (7.8%) of the sets were VRE positive in at least one of five samples. The negative predictive value of van qPCR on the first enrichment broth was 99.3%. With a cutoff quantification cycle (C_q) value of >35 to discriminate negative and positive samples, 87% of the negative patients can be identified within a day after obtaining the sample, compared to 7 days in the culturing approach. VRE screening using qPCR on one enrichment broth can quickly identify non-VRE-colonized patients and therefore decrease costs and limit unnecessary isolation restrictions.

INTRODUCTION

In Europe, the rate of vancomycin-resistant Enterococcus (VRE) colonization on intensive care unit (ICU) admission is estimated at 2.7%, compared to 12.3% in the United States (1). However, VRE prevalence increased significantly in Europe between 2011 and 2014, from 6.2% to 7.9% (2). Although VRE prevalence in The Netherlands is low at 0.9%, 16 VRE hospital outbreaks have been reported to the Signaling Consultation of Hospital acquired Infections and AntiMicrobial Resistance (SO-ZI/AMR) in 2015 (3). Infection control practices for VRE differ globally, but most hospitals give patients who are VRE colonized a VRE flag (4), and this flag stays with the patient upon hospital dismissal until the criteria for removing the flag are met. In The Netherlands, it is common practice to preemptively flag patients who have been in contact with known VRE-colonized patients in an outbreak situation (5). Large outbreaks can result in thousands of patients being flagged, and since contact isolation precautions are taken when flagged patients are hospitalized, isolation room capacity can become insufficient. It is therefore important to optimize VRE screening methods to reduce the time between sampling a patient and that of establishing the colonization status of the patient (currently 7 days in culturing sets of five rectal samples).

There is no consensus on the best diagnostic screening protocol to determine VRE colonization, and methods differ greatly between different studies (6). In 2015, a Dutch guideline was published which states that three to five rectal swabs should be taken, on separate days, in screening VRE-flagged patients for VRE (7); procedures to follow up on previously VRE-colonized individuals are less formalized in terms of the frequency and number of samples obtained for screening.

Our local hospitals currently follow those guidelines on VRE testing (7), whereby after incubation in an enrichment broth, samples are directly cultured using a selective (chromogenic) agar. If the first rectal sample is taken on day 0, rectal samples are taken daily until day 4, and, finally, the reading of the last agar plate for a VRE-negative patient is done on days 6 and 7 (VRE-positive samples are known as soon as any of the five plates is found to be positive). In contrast, the result of a screening quantitative PCR (qPCR) performed on the enrichment broth of the first sample could identify VRE status on day 1. We therefore wanted to establish the negative predictive value of a van qPCR on the first enrichment broth, assuming no ongoing VRE exposure (i.e., assuming a stay in a VRE-free hospital or home situation). This is relevant for VRE-negative patients, as absence of VRE colonization would remove the need for precautionary isolation measures, and they would be able to be moved to the ward upon establishing negative qPCR results on day 1 with qPCR rather than on day 7 with culturing. Targeted antimicrobial use and the need for fewer swabs would further improve patient care and well-being. Costs would also be reduced by avoiding unnecessary treatment in isolation and extensive culturing of negative samples. Finally, faster diagnostic assessment and the concomitant better insights into epidemic spread would enable a more adequate and rapid response in the management of an ongoing outbreak, such as separating noncolonized from colonized patients, flagging patients, and closing down wards if necessary.

RESULTS

We analyzed 255 sets of five rectal specimens from 243 patients obtained between December 2014 and June 2016 (see Table 1). A total of 84 of these specimens were from known VRE carriers, with an average time since their most recent VRE-positive test of 455 days (range, 72 to 779 days), and 171 sets were from patients whose VRE colonization status was unknown. Twenty (7.8%) of these 255 sample sets were VRE positive in culture, which is the current hospital practice and the reference standard in our study.

TABLE 1.

Age, sex, and culturing results for the group of known VRE carriers, for individuals with unknown VRE status, and for the total population

Parameter	Value(s)
	Known VRE carriers (n = 84)	Unknown VRE status (n = 171)	Total population (n = 255)
Sex
No. of males	50	86	136
No. of females	34	85	119
Mean patient age in yrs (range)	72.5 (47–91)	66.1 (17–95)	68.2 (17–95)
No. of culture-positive results	10	10	20

We performed van qPCR on the first enrichment broth of each set, and 109 (43%) of the 255 samples were positive in the van qPCR (vanA, n = 5; vanB, n = 101; vanA and vanB [vanA/B], n = 3; see Table 2). Of the culture-positive samples, two were vanA positive in qPCR and 17 were vanB positive in qPCR. One culture-positive set was qPCR negative, as the first enrichment broth had no detectable quantification cycle (C_q) value. Interestingly, the broths of the remaining four samples were also measured and had C_q values ranging between 21 and 25. Moreover, while the enrichment broth of this first swab was also culture negative, the next broth was culture positive (further broth cultures were not performed upon finding a positive culture).

TABLE 2.

Confusion matrix of qPCR versus culturing results

Result	No. of results
	Culture positivea	Culture negativeb	Total
Detected by PCRc	19	90	109
Not detected by PCRd	1	145	146
Total	20	235	255

^aSensitivity, 95.0% (95% confidence interval, 75% to 100%).

^bSpecificity, 61.7% (95% confidence interval, 55% to 68%).

^cPositive predictive value (PPV), 17.4% (95% confidence interval, 11% to 26%).

^dNegative predictive value (NPV), 99.3% (95% confidence interval, 96% to 100%).

Table 2 shows that the negative predictive value of van qPCR on the first enrichment broth for predicting a VRE-negative set of five culturing specimens was 99.3% (confidence interval [CI], 96.4% to 100%), with a specificity of 61.7%. Figure 1A displays the measured C_q values for vanA and vanB and indicates whether the culture results were negative or positive. Of the 235 culture-negative samples, the data from 145 did not include detectable C_q values.

FIG 1.

Discriminating possible VRE-positive samples on the basis of C_q values. (A) The qPCR results were sorted by C_q values. One positive and 145 negative samples were not detected (n.d.) by qPCR. (B) Sensitivity and specificity as a function of C_q threshold using the culture results as the gold standard.

The qPCR-positive VRE culture-positive sets had C_q values between 17 and 34 upon examination of the first enrichment broth; the culture-negative sets with detectable qPCR signal had C_q values between 21 and 45. Of the 11 culture-negative sample sets with C_q values below 30 (all vanB), 7 sets were from known VRE carriers. Five of these seven known carriers had had over 10 negative VRE cultures since the last VRE detection. Tables S1 and S2 in the supplemental material show the confusion matrices comparing the qPCR and culturing results separately for the group of known VRE carriers and for the group with unknown VRE status. No significant differences between these two groups were observed, although there was higher specificity seen in the group of patients with unknown VRE status than in the known VRE carriers.

We then evaluated the effect of the use of a qPCR cutoff C_q value of >35, a conservative threshold based on the highest C_q value determined for a culture-positive sample (C_q = 33.95). With the use of this C_q threshold value of >35 to discriminate culture-negative sample sets, the specificity of qPCR was 87% and the sensitivity was 95% (see Fig. 1B and Table 3). This means that only 13% of the VRE-negative samples needed further culture testing based on a positive qPCR screening result. In evaluating the known VRE carrier group and the patient group with unknown VRE status separately, no significant differences were detected, although the specificity was again higher in the group with unknown VRE status (see Tables S3 and S4).

TABLE 3.

Confusion matrix of qPCR C_q versus culturing results

Resulta	No. of results
	Culture positiveb	Culture negativec	Total
Cq ≤ 35d	19	31	50
Cq > 35 or NDe	1	204	205
Total	20	235	255

^aThe qPCR groups are separated into a group with C_q values of ≤35 and a combined group consisting of those with C_q values of >35 plus those that were not detected (ND).

^bSensitivity, 95.0% (95% confidence interval, 75% to 100%).

^cSpecificity, 86.8% (95% confidence interval, 82% to 91%).

^dPPV, 38.0% (95% confidence interval, 25% to 53%).

^eNPV, 99.5% (95% confidence interval, 97% to 100%).

qPCR was performed on all five enrichment broths for 50 sample sets, and one or more of the qPCRs were positive in 31 sets. The spread in the C_q values for these 31 sets with at least one detectable C_q value is illustrated in Fig. 2. Of these 50 sample sets, 8 were VRE positive (2 vanA and 6 vanB): for the vanB-positive samples, the maximum difference between the minimum and maximum C_q values was 12 (between C_q 19.2 and C_q 31.2), and the average of the range is 6.8. In one set, although the first broth had a C_q value below 35, one of the four subsequent broths had a C_q value of only 37.5. One vanA culture-positive patient had four negative qPCRs, while one qPCR had a C_q value of 25.4. In another vanA culture-positive patient also described above, one qPCR was negative, while the C_q for the other four samples ranged between 21.1 and 24.6.

FIG 2.

Variability in C_q values. For 50 sample sets, we measured the qPCR results for all five broths in the sample series. In 31 of these sets, at least one of the samples had a detectable qPCR signal (n.d. = not detected). For one sample which was culture negative, both a vanA signal and a vanB signal were detected, both results are displayed. The mean of the results from the 5 samples is displayed with a black circle; in cases in which 1 of the 5 samples was negative, the mean of the positive samples was calculated and the result is displayed with an open circle. To arrange samples along the horizontal axis, samples were sorted according to their mean C_q values (if not determined, the value was set to 46.5 in calculating the mean), in groups of culture-positive (split between vanA and vanB results) and culture-negative samples.

DISCUSSION

Since the incidence of VRE and, consequently, the number of individuals with VRE flags assigned as a consequence of previous colonization or possible colonization, by being categorized as at risk, are increasing, fast and accurate methods of screening upon hospital admission of VRE-flagged patients are needed. Our results show that one van qPCR performed on enrichment broth can replace five rectal swab cultures if the van qPCR result is negative. In our study, 1 of 20 positive culture sets was negative in the qPCR result from the first broth but showed high C_q values in the four other broths. We therefore hypothesize that this was caused by sampling error rather than by a lack of sensitivity in the qPCR methodology. Omitting this possible sampling error, the sensitivity of qPCR in our study was 100%. Sampling error is also likely to have occurred in the set of five specimens where four negative qPCRs were accompanied by a qPCR with a C_q value of 25.4—especially in the context of the limited qPCR variability observed throughout the rest of our study results as indicated in Fig. 2.

Using VRE culturing of a sample set of five specimens as the gold standard, the negative predictive value of the van qPCR performed on enrichment broth was 99.3%. Performing van qPCR on enrichment broth without a C_q cutoff value made VRE culturing superfluous for 57% of our sets. Using a qPCR cutoff value of >35 to discriminate unculturable sample sets, 80% of our sets were correctly identified as negative results without decreasing sensitivity. Of course, the relative benefits of the screening further depend on the incidence of VRE—in our study, we preferentially selected VRE-positive samples, thereby effectively also decreasing our negative predictive value and increasing our positive predictive value.

The use of one qPCR instead of five cultures results in a shorter turnaround time and is more patient friendly and reduces work load and costs by decreasing the number of isolation days. Moreover, it would enable hospitals to quickly separate VRE-colonized patients from noncolonized patients after VRE flagging of large groups of hospitalized patients during an outbreak, which could help in containing the outbreak. Removal of patient flags also reduces the workload, which could result in higher rates of compliance with infection control measures in other patients.

Our study had some limitations, for example, the need to refer to culture data as the gold standard. There was a large group of sample sets that were qPCR positive, or had a detectable qPCR signal, while being negative in culture. There are at least two alternative explanations for this. First, the combination of a detectable C_q value with a negative culture result could be due to the detection of vanB genes in anaerobes (8). However, the use of amoxicillin in the enrichment broths kills most anaerobes. Second, a low sensitivity in VRE culturing could cause this phenomenon. In such a case, the VRE prevalence in our population could be underestimated. This is compatible with our observation that some previously VRE-colonized patients had multiple negative VRE cultures while having low C_q values in the enrichment broth qPCR, suggesting that these patients were VRE colonized but that culturing was not sensitive enough. It is also in line with the lower specificity observed in the group with known VRE history than in the group with unknown VRE status. D'Agata et al. also reported that VRE screening using rectal swabs instead of stool samples has limited sensitivity (9), but the approach is widely used for practical reasons. This scenario of potentially undetected VRE-colonized patients should be studied further as it is of great interest for infection control management—and could be alleviated by using the more sensitive qPCR approach rather than culturing.

In our study, using a cutoff C_q value of >35, only one false-negative qPCR result for the first broth occurred and was probably due to sampling error. Therefore, this cutoff value is appropriate for our suggested protocol despite the variation in C_q values within sample sets, and we have implemented this threshold for practical use in screening for infection control. In a previous study by Zhou et al., qPCR on enrichment broth was used on a single rectal swab (10). Using C_q cutoff values of ≤25 and >30, this study separated true positive and negative samples based on culture data. We were not able to set a C_q cutoff value for definitive positive samples due to the variation in C_q values in VRE-positive and -negative sample sets.

Our data show that analyzing one rectal sample with qPCR efficiently screens for negative samples prior to culturing of five rectal specimens. One caveat concerns the larger effect of sampling error such as was the case for one patient in our study, since only one sample was subjected to qPCR compared to five for culturing. In our hospitals, we therefore implemented the protocol that someone who was VRE flagged but not known to be VRE positive requires one negative qPCR whereas a known previous VRE carrier requires at least two negative VRE qPCR results before removal of the patient flag. Also, the cutoff C_q value of >35 suggested in our study should be validated for every laboratory individually since detection limits can vary, e.g., with different primers and probes or other extraction and qPCR methodologies. In our setting, the protocol outlined in this study has successfully reduced the pressure on isolation capacity by clearing the VRE flag for 87% of VRE-negative patients in 1 day instead of 7 days, decreasing costs and improving patient care and experience.

MATERIALS AND METHODS

Sample selection.

During and following a large VRE outbreak that started in 2013 in the Eindhoven area (The Netherlands), we received rectal specimens for culture testing: as part of the outbreak and follow-up investigation, confirmed and exposed patients were invited to submit a series of five samples obtained on consecutive days. The patient group therefore consisted of both hospitalized patients and patients who were invited for follow-up testing via home sampling. For our analysis, we retrospectively performed qPCR on a subset of these sample sets, obtained between December 2014 and June 2016. We included only cases where the culture was positive or all five rectal specimens were available. Although not all VRE-negative patients were included as they were well represented, we ensured that VRE-positive sample sets were included in order to test the performance of qPCR versus that of culturing. This resulted in 255 sets of five rectal specimens from 243 patients.

The study protocol was reviewed and approved by the medical ethics board of the Máxima Medical Center (study number N16.044). The requirement for informed consent was waived because only residual material was used.

VRE culture testing.

Rectal samples were obtained and eluted into medium using ESwab (Copan). Subsequently, a 0.5-ml sample was transferred and cultured for 22 h using a brain heart infusion enrichment broth with 16 mg/liter amoxicillin at 35°C under standard atmospheric conditions. The sample was then incubated on chromogenic agar (Brilliance VRE agar; Thermo Fisher Scientific) for 48 h at 35°C (standard atmospheric conditions). Suspected colonies were inoculated on a blood agar plate for 18 to 24 h at 35°C under standard atmospheric conditions. Colonies (from blood agar) were identified by matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) (Bruker) followed by a vancomycin Etest (bioMérieux) if Enterococcus spp. were detected. If the vancomycin MIC was >4 mg/liter or the borders of the vancomycin Etest were fuzzy, the strain was analyzed with van qPCR and an amoxicillin Etest (bioMérieux).

VRE qPCR testing.

All enrichment broths of the first rectal swab (incubated for 22 h) were also analyzed with van qPCR. DNA was isolated from 100 μl broth using a Versant sample preparation 1.0 kit on a kPCR sample preparation (SP) module (Siemens). We performed vanA/B qPCR using 20 μl of PCR mix containing 300 nM primers and 200 nM probes and 4× TaqMan Fast virus 1-Step master mix (Thermo Fisher Scientific) and 5 μl DNA eluate. Primers and probes have been described by Palladino et al. (11). Cycling conditions were 55°C for 5 min, 95°C for 20 s, and 45 cycles of 95°C for 3 s and 60°C for 30 s. Samples were classified as VRE positive if enterococci were amoxicillin resistant and the van qPCR result was positive. To investigate the variance in C_q values, qPCR was also performed on a subset of the second to fifth enrichment broths in a subset of 50 sample sets.

Statistical analyses.

A patient was considered VRE positive if VRE was detected in at least one of five cultured specimens (the gold standard). Using a confusion matrix of culture-positive/culture-negative status and qPCR-positive/qPCR-negative status as a function of the C_q threshold value, we calculated the sensitivity and specificity values shown in Fig. 1B.

The epiR package was used to calculate the 95% confidence intervals for the sensitivity and specificity and positive and negative predictive values (R version 3.3.2; The R Foundation for Statistical Computing, Vienna, Austria).