Reduction of the use of antimicrobial drugs following the rapid detection of Streptococcus agalactiae in the vagina at delivery by real-time PCR assay

Authors

  • E Poncelet-Jasserand,

    1. Department of Obstetrics, Gynaecology and Reproductive Medicine, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
    • These authors contributed equally to the work.

  • F Forges,

    1. Unit of Clinical Research, Innovation and Pharmacology, INSERM CIE3, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
    • These authors contributed equally to the work.

  • M-N Varlet,

    1. Department of Obstetrics, Gynaecology and Reproductive Medicine, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
  • C Chauleur,

    1. Department of Obstetrics, Gynaecology and Reproductive Medicine, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
  • P Seffert,

    1. Department of Obstetrics, Gynaecology and Reproductive Medicine, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
  • C Siani,

    1. Research Laboratory in Knowledge Engineering (ERIC), University of Lyon, Institute of Pharmaceutical and Biological Sciences (ISPB) of Lyon, France
    Search for more papers by this author
  • B Pozzetto,

    Corresponding author
    1. Groupe Immunité des Muqueuses et Agents Pathogènes (GIMAP), Faculty of Medicine of Saint-Etienne, University of Lyon, France
    2. Department of Microbiology, Hôpital Nord, University-Hospital of Saint-Etienne, France
    • Department of Obstetrics, Gynaecology and Reproductive Medicine, University-Hospital of Saint-Etienne, France
    Search for more papers by this author
  • A Ros

    1. Department of Microbiology, Hôpital Nord, University-Hospital of Saint-Etienne, France
    Search for more papers by this author

Correspondence: Prof Bruno Pozzetto, GIMAP EA 3064, Faculté de Médecine J. Lisfranc, 15 rue Ambroise Paré, 42023 Saint-Etienne Cedex 02, France. Email bruno.pozzetto@univ-st-etienne.fr

Abstract

Objective

To assess whether the determination of the presence of group B streptococci (GBS) in the vagina using a rapid polymerase chain reaction (PCR) assay at delivery was able to spare useless antimicrobial treatments, as compared with conventional culture at 34–38 weeks of gestation.

Design

Practical evaluation and prospective cost-effectiveness analysis.

Setting

A university hospital in France.

Population

A cohort of 225 women in labour at the University-Hospital of Saint-Etienne.

Methods

Each woman had a conventional culture performed at 34–38 weeks of gestation. At the beginning of labour, two vaginal swabs were sampled for rapid PCR testing and culture. The decision to prescribe a prophylactic antimicrobial treatment or not was taken according to the result of the PCR test. A comparative cost-effectiveness analysis of the two diagnostic strategies was carried out.

Main outcome measures

Number of women receiving inadequate prophylactic antimicrobial drugs following each testing strategy, costs of PCR testing and culture, frequency of vaginal GBS, and diagnostic performance of the PCR test at delivery.

Results

The percentage of unnecessarily treated women was significantly reduced using the rapid test versus conventional culture (4.5 and 13.6%, respectively; P < 0.001). The rate of vaginal GBS at delivery was 12.5%. The incremental cost-effectiveness ratio (ICER) for each inadequate management avoided was €36 and €173 from the point of view of the healthcare system and hospital, respectively.

Conclusions

The PCR assay reduced the number of inadequate antimicrobial treatments aimed to prevent the early onset of GBS disease. However, this strategy generates extra costs that must be put into balance with its clinical benefits.

Introduction

Streptococcus agalactiae (group B streptococci, GBS) is a leading cause of neonatal infections in developed countries: it is responsible for pneumonia, septicaemia, and meningitis, which can become life-threatening in neonates.[1] The bacterium is acquired by contact with the mother's vaginal epithelium during delivery. The rate of vaginal colonisation ranges from 11 to 21% in Western Europe.[2, 3] In 2001, to reduce the risk of GBS neonatal infection, experts from the Agence Nationale d'Accréditation et d'Evaluation en Santé (ANAES) recommended the implementation of the systematic detection of vaginal colonisation by this bacterium through culturing a vaginal sample during the final weeks of gestation (usually at 34–38 weeks of gestation). A prophylactic antimicrobial treatment administered during the course of the delivery has been shown to prevent neonatal infection.[4] It usually consists of amoxicillin (an initial infusion of 2 grams, followed by an infusion of 1 g every 4 hours until delivery), which can be replaced by erythromycin or clindamycin in the case of allergy to beta-lactams. This strategy has significantly reduced the incidence of early neonatal infections, from 1.5 per 1000 births at the end of the 1980s to 0.37 per 1000 births in 2004.[5, 6] Two main reasons can explain the remaining cases of GBS neonatal infections: first, at least 15% of pregnant women are not currently tested for GBS vaginal colonisation,[7, 8] and second, women tested negative at 34–38 weeks of gestation could be positive at delivery as a result of an intermittent vaginal colonisation.[9] Indeed, more than 70% of early GBS neonatal cases of bacteraemia or meningitis analysed at the French National Reference Laboratory of Streptococci in Paris corresponded to an unknown or negative colonisation status for the mother (C. Poyard, pers. comm.).

Molecular tests that allow the rapid screening of genital colonisation have been developed.[10-13] In a meta-analysis,[14] their sensitivity and specificity ranged from 0.94 to 0.97 (95% CI 0.95–1.00) and 0.96 to 1.00 (95% CI 0.94–1.00), respectively. One test, GenExpert,™ manufactured by Cepheid (Maurens-Scopont, France), is very simple to use and can be performed in less than 3 hours: it was evaluated positively for the detection of GBS vaginal colonisation in pregnant women.[3, 15-18] In a French study comparing this test with the conventional detection of vaginal colonisation at 35–37 weeks of gestation, the intermittent presence of vaginal GBS between 35–37 weeks of gestation and delivery was observed in 12.1% of the 933 women tested, with the test being positive and negative in 7.0 and 5.1% of women, respectively.[17] In the latter situation, prophylaxis using an antimicrobial drug is useless, with a possible adverse impact on the susceptibility of GBS to antibiotics. Presently in France the molecular detection of GBS by rapid techniques is only recommended in women not tested for vaginal colonisation at 34–38 weeks of gestation and hospitalised for early preterm prelabour rupture of membranes, thereby excluding women with intermittent vaginal GBS.[9]

Costs and cost-effectiveness ratios of rapid testing at delivery have already been estimated within the National Health Service in the UK.[19] In this review, comparing several strategies, polymerase chain reaction (PCR) did not appear to be a cost-effective strategy in comparison with screening at 34–38 weeks of gestation; however, the Cepheid rapid PCR assay was not evaluated. Very recently, El Helali et al. reported a before-and-after study comparing 2761 term deliveries in 2009 screened for GBS vaginal colonisation at 34–38 weeks of gestation with conventional technique and 2814 term deliveries in 2010 screened at delivery for the same parameter with the Cepheid test. They concluded that the screening by PCR at delivery reduced the number and severity of early-onset GBS disease, and also reduced the hospital costs incurred.[20]

The primary objective of the present study was to assess whether the determination of vaginal GBS by using the Cepheid rapid PCR assay at delivery was able to spare useless antimicrobial treatments, in comparison with conventional culture at 34–38 weeks of gestation. In parallel, the practical feasibility of the test being available 24 hours a day, 7 days a week, was analysed, together with its analytical performance with reference to bacterial culture. The costs of this new strategy were also assessed in comparison with the screening at 34–38 weeks of gestation, as recommended by French guidelines.[4]

Methods

Pregnant women

All pregnant women who attended the Department of Obstetrics at the University-Hospital of Saint-Etienne for delivery were asked to participate in the study after they received oral and written information, with the exception of women under 18 years of age, women under guardianship, women who delivered anonymously, women with an indication for scheduled caesarean section, women presenting with signs of imminent delivery (defined as a cervical dilation > 5 cm), women with a pregnancy of under 35 weeks of gestation, women who received an antimicrobial treatment against GBS for less than 1 week, and women who refused to participate. The protocol was submitted for the approval of the ethics committee of the University-Hospital of Saint-Etienne.

Evaluation of the population size

According to El Helali et al., 5% of pregnant women test negative for vaginal GBS at delivery, but test positive in the screening at 34–38 weeks of gestation, and thus may receive pointless prophylactic treatment.[17] Furthermore, according to previous studies,[7, 8] approximately 15% of the pregnant women still miss the screening at 34–38 weeks of gestation; thus, assuming a prevalence of GBS of 12.3%, 13.2% of the women with unknown status also receive useless prophylaxis. Altogether, 18.2% of women may receive an inadequate prophylaxis treatment. With the objective of a 50% reduction of the relative risk, with an alpha risk of 5% and a power of 80%, at least 220 pregnant women would be required to test the new protocol, considering each woman as her own control.

Microbiological methods

Screening by real-time (RT) PCR at delivery was performed by the RapidXpert GBS™ test, aimed to detect GBS DNA on GenExpert,™ Cepheid, on a routine basis, 24 hours a day, 7 days a week, by the technicians of the microbiology laboratory during opening hours, and by residents specialised in biology during nights and weekends. The sensitivity of the PCR assay specified in the handbook of the test was 620 units forming colony per swab. The screening test at 34–38 weeks of gestation and the intrapartum control test at delivery were performed by plating a vaginal swab on Columbia colistin and nalidixic acid and Granada media incubated at 37°C, under 5% CO2, for the first medium, and under anaerobic conditions for the second.

Design of the study

When the pregnant women arrived at hospital for delivery, they were asked to enter the study. Each woman was her own control. Two vaginal swabs were recovered, one for agar culture and the other for PCR. The decision to prescribe a prophylactic antimicrobial treatment was taken according to the result of the PCR test. In the case of a positive or non-interpretable result, the antimicrobial treatment was performed according to the current recommendations (infusion of 2 g of amoxicillin, followed by an infusion of 1 g every 4 hours until delivery, or a single infusion of 1.2 g of clindamycin in the case of allergy to beta-lactams), and complementary bacterial specimens were sampled from the neonate. In the case of a negative result, no treatment was given.

The parameters systematically collected for each woman were as follows: result of vaginal GBS screening at 34–38 weeks of gestation (negative, positive, or not determined); duration of pregnancy (weeks of gestation); previous episodes of vaginal GBS; if available, the exact time at which the samples were sent to the laboratory and at which the PCR result was communicated to the obstetrician; antimicrobial treatments received by the mother and her baby; additional samples performed in the neonate to document sepsis; and follow-up of the neonate during the days after birth.

Cost-effectiveness analysis

For the economic analysis, the following resources were taken into consideration as relevant variable costs: microbiological tests used for the detection of vaginal GBS; antimicrobial regimens; and additional biological tests. Costs were calculated according to the French listing of biological tests for women included in this list, and for the data available from the Pharmacy Department of our Hospital for antimicrobial drugs and medical devices. The cost of the PCR test was estimated according to the list of non-registered acts available on the website of the University-Hospital of Montpellier, matched with the innovative biological technologies, the costs of which are not yet fixed by the French authorities. The differential costs related to nursing staff (supervision of presumably infected babies) or to technician staff were considered as fixed costs, and were thus not taken into account, because it was assumed that this new strategy would be performed without personnel reduction. Other fixed costs were similar in the two groups, and thus were not assessed. Two points of view were considered: that of the French healthcare system, which funds the whole antepartum screening programme (in-hospital and private GBS screening), and that of the hospital, which only funds the in-hospital GBS screening. The number of inadequate treatments avoided (i.e. women who were wrongly treated or wrongly not treated) was considered as the clinical effectiveness outcome. The incremental cost-effectiveness ratio (ICER) was expressed as cost per avoided inadequate treatment during 1 year, and is defined as: ICER = (costPCR – costculture)/(effectivenessPCR – effectivenessculture).

Statistical analysis

The diagnostic performance (sensitivity, specificity, predictive values, and relevance of prophylactic treatments) was estimated for the two diagnostic strategies. A woman with an unavailable screening result was considered as GBS positive, and received a prophylactic treatment. The technical performance of the PCR assay was also described.

The costs were estimated by the microcosting method for the PCR results. The PCR outcomes were compared with the reference decision tree, taking into account the culture results at 34–38 weeks of gestation. The McNemar test was used to compare qualitative paired data. Costs and time responses were described with means, medians, and quartiles. To account for uncertainty because of sampling fluctuations, confidence regions around the ICER were calculated using the truncated Fieller's method.[21-23] A graphical representation of this uncertainty around the ICER on the cost-effectiveness (CE) plane was based on a non-parametric bootstrap method.[24] For greater robustness in the results, uncertainty around the ICER was also taken into account by providing probabilities that the true ICER belonged to each of the quadrants of the CE plane.

Finally, to account for uncertainty from the various model hypotheses on parameters, a sensitivity analysis was also conducted by varying the PCR cost, which is the only uncertain parameter. Five scenarios for the unit cost of PCR testing were considered from the healthcare system point of view, which would be the more relevant decision maker.

The statistical analyses were performed using sas 9.2 (Cary, NC, USA) and gauss® 9.0 (Aptech Systems Inc., Black Diamond, WA, USA). A threshold of 5% was chosen for significant results.

Results

Management of antimicrobial prophylaxis according to the strategy used for detection of vaginal GBS

From a total of 362 deliveries that occurred from 19 November 2008 to 2 January 2009, 225 women were prospectively and consecutively included in the study according to the criteria defined above (62.1%). The distribution of excluded cases was as follows: 54 scheduled caesarean sections (14.9%), 40 women with cervical dilation over 5 cm (11.1%), and 43 cases with miscellaneous reasons (antibiotic treatment against GBS for less 1 week, pregnancy duration under 35 weeks of gestation, or refusal to participate; 11.9%).

The results of the strategies using PCR at delivery and culture at 34–38 weeks of gestation for the prediction of vaginal GBS at delivery in the 225 women included are depicted in Figures 1 and 2, respectively. The percentage of women who received an unnecessary antimicrobial regimen was 4.5 and 13.6% using PCR assay at delivery and culture at 34–38 weeks of gestation, respectively (< 0.001 by McNemar test).

Figure 1.

Results of the detection of vaginal group B streptococci (GBS) by bacterial culture at 34–38 weeks of gestation (WG), by comparison with bacterial culture at delivery taken as the gold standard. *Women from this group were considered positive. NA: data not available.

Figure 2.

Results of the detection of vaginal group B streptococci (GBS) by GenExpert PCR at delivery, by comparison with bacterial culture at delivery taken as the gold standard. *Women from this group were considered negative, as no prophylactic treatment was administered. **Women from this group were considered positive. NA: data not available.

All the other performance parameters were improved by using the PCR assay (Table 1). In the group of 18 women who received adequate prophylactic antibiotic treatment, and who were considered as truly positive for GBS colonisation (Figure 2), ten received a correct dose of antibiotics (a total of 3 grams of amoxicillin or 1.2 g of clindamycin), and eight received just 2 g of amoxicillin because their labour lasted less than 6 hours. From the latter group, five women delivered in less than 4 hours, which means that they would not have received a sufficient dose of amoxicillin even if the result of the GBS test had been available at the time of arrival at hospital. Finally, only three of these 18 women (16.7%) received just 2 grams of amoxicillin because of the delay generated by the PCR test.

Table 1. Performance of each strategy of detection for vaginal group B streptococci (GBS), with bacterial culture at delivery taken as the gold standard
 Culture at 34–38 weeks of gestationPCR at deliveryStatistical analysisa
  1. a

    McNemar test.

Diagnostic performance (%):
Sensitivity (95% CI)55.6 (35.3–74.5)66.7 (46.0–83.5) 
Specificity (95% CI)84.5 (78.7–89.3)94.9 (90.9–97.5) 
Positive predictive value (95% CI)33.3 (20.0–48.9)64.3 (44.1–81.4) 
Negative predictive value (95% CI)93.2 (88.4–96.4)95.4 (91.5–97.9) 
Women inadequately treated with prophylactic antimicrobial treatment (%)13.64.5< 0.001
Women adequately treated with prophylactic antimicrobial treatment (%)6.88.0NS
Women inadequately not treated with prophylactic antimicrobial treatment (%)5.44.0NS
Women adequately not treated with prophylactic antimicrobial treatment (%)74.283.5< 0.05
Technical performances (%):
Sensitivity (95% CI)58.3 (36.6–77.9)76.9 (56.3–91.0) 
Specificity (95% CI)92.7 (87.8–96.0)95.4 (91.4–97.9) 
Positive predictive value (95% CI)51.9 (31.9–71.3)69.0 (49.2–84.7) 
Negative predictive value (95% CI)94.3 (89.7–97.2)96.9 (93.3–98.8) 

Performance of the PCR assay as compared with retrospective culture

After a training period of a few days, only two tests could not be interpreted (0.89%). The median response time for PCR results in the ward was 1 hour 50 minutes (ranging from 1 hour 35 minutes to 2 hours 20 minutes). Figure 3 illustrates the distribution of the times between admission and when the PCR result was available in the ward. The diagnostic (Table 1) and technical (Tables 1 and 2) performances of the PCR assay were systematically favourable in comparison with those of the culture test. Indeed, the PCR test was more sensitive (66.7 versus 55.6%), more specific (94.9 versus 84.5%), and exhibited better positive and negative predictive values (64.3 versus 33.3%, and 95.4 versus 93.2%, respectively) than the culture.

Table 2. Evaluation of PCR at delivery and bacterial culture at 34–38 weeks of gestation, with reference to bacterial culture at delivery taken as the gold standard for the detection of vaginal group B streptococci (GBS)
 Culture at deliveryTotal
PositiveNegative
  1. a

    For one woman, some data were missing.

  2. b

    For four women, some data were missing.

PCR at delivery
Positive or not interpretable181028
Negative or with result available only after delivery9187196
Total27197224a
Culture at 34–38 weeks of gestation
Positive or not available153045
Negative12164176
Total27194221b
Figure 3.

Histogram of the distribution of times between admission and the result of the group B streptococci screening in vaginal secretions by PCR being available in the ward for the 225 women included into the study.

Epidemiological characteristics of the pregnant women

The epidemiological data are listed in Table 3: the mean time duration of pregnancy was 39.4 weeks of gestation (95% CI 39.2–39.6 weeks of gestation). From the 225 women included in the study, the screening culture at 34–38 weeks of gestation was lacking in 8.1%. Vaginal GBS was found in 12.5% of women at delivery versus 12.6% at 34–38 weeks of gestation. In the group of women with results available from screening at 34–38 weeks of gestation, 12.3% had their vaginal GBS status change at delivery: 5.9% of them changed from negative to positive, whereas 6.4% of them changed from positive to negative. Importantly, no neonatal infection linked to GBS was recorded during the study.

Table 3. Epidemiological characteristics of vaginal group B streptococci (GBS) in the 225 pregnant women included in the study
ParameterNo.Total(%)
  1. Some data were missing.

Notion of prior vaginal GBS6216(2.7)
Prevalence of vaginal GBS by culture at 34–38 weeks of gestation28222(12.6)
Prevalence of vaginal GBS by culture at delivery28224(12.5)
Prevalence of vaginal GBS by PCR at delivery29225(12.9)
Loss of vaginal GBS between 34–38 weeks of gestation and delivery13203(6.4)
Acquisition of vaginal GBS between 34–38 weeks of gestation and delivery12203(5.9)
Occurrence of delivery prior to the availability of the PCR result20201(9.1)
Unknown status of vaginal GBS at delivery18222(8.1)
PCR result not interpretable2225(0.9)

Results of the cost-effectiveness analysis

The results of the economic analysis are shown in Table 4. Whatever the point of view chosen, and because the PCR costs were supported by the hospital, the economic analysis showed a mean cost per patient of €69.40 and a median cost per patient of €40.5 (range: €40.50–125.10) for the PCR test at delivery. On the other hand, the costs of the antepartum culture differed depending on the adopted point of view. Indeed, only 54% of the antepartum cultures were performed by the hospital laboratory; the other 46% were either performed by private laboratories or not performed at all. Thus, from the healthcare system point of view, the economic analysis showed a mean cost per patient of €65.60 and a median cost per patient of €37.80 (range: €37.80–103.30) for screening by culture at 34–38 weeks of gestation. From the hospital point of view, the mean cost per patient was €51.40 and the median cost per patient was €37.80 (range: €0.00–93.10) for screening by culture at 34–38 weeks of gestation. On the basis of 3200 delivery acts performed each year in our hospital, the extra cost of the PCR strategy would be €57 719 for the hospital setting and only €12 084 for the healthcare system. According to our results, the strategy using preterm culture for the assessment of vaginal GBS led to the inadequate management of 19% of the pregnant women, versus only 8.5% for the strategy using the intrapartum PCR test, which means that during one year, respectively, 596 and 263 pregnant women would be inadequately managed.

Table 4. Results of the cost-effectiveness analysis comparing antepartum culture and intrapartum PCR
 Antepartum cultureIntrapartum PCR
  1. Q, quartile.

Mean cost (SD) (€/patient)
Hospital51.4 (45.50)69.4 (42.00)
Healthcare system65.6 (40.00)
Median cost (Q1–Q3) (€/patient)
Hospital37.8 (0.00–93.10)40.5 (40.50–65.10)
Healthcare system37.8 (37.80–103.30)
Cost (SD) for 3200 patients (€/year)
Hospital164 359 (145 523)222 078 (133 960)
Healthcare system209 994 (127 948)
Effectiveness (SD) (number of inadequate managements/year) 596 (1249)263 (881)
Difference of costs (SD) (€/year)
Hospital 57 719 (13 380)
Healthcare system 12 084 (12 532)
Difference of effectiveness (SD) (inadequate management/year) −333 (103)
ICER (95% CI) (€/avoided inadequate management)
Hospital173 (72.00; 524.60)
Healthcare system36 (−34.00; 190.00)

The ICER per avoided inadequate management was €36 and €173 from the point of view of the healthcare system and hospital, respectively. The results are reported in Table 4.

The graphical representation of the uncertainty around the ICER on the CE plane is shown in Figure 4. As the origin of the cost-effectiveness plane was not included in the inner 95% confidence ellipse, rapid PCR testing at delivery versus conventional culture at 34–38 weeks of gestation is significantly different in terms of cost and effectiveness.

Figure 4.

Probabilistic analysis of the incremental cost-effectiveness ratio (ICER) for the hospital (A) and the healthcare system (B). inline image denotes the difference in average total costs. inline image denotes the difference in average effectiveness. The scatter of points corresponds to 10 000 non-parametric bootstrap replications of the pair (inline image; inline image). In each quadrant, the percentage quantifies the probability that the true ICER is in the quadrant under consideration. The outer ellipse defines the confidence region of the pair (inline image; inline image) at a level of 95%, and the inner ellipse defines that at a level of 50%.

The probability of the ICER belonging to each quadrant of the CE plane, presented in Figure 4, is highest for the north-east quadrant (83.2% from the healthcare system point of view; 99.9% from the hospital point of view), in which rapid PCR testing at delivery is both more costly and more effective than conventional culture at 34–38 weeks of gestation, but with a reasonable upper limit for the 95% confidence region of the ICER (€190.00 per avoided inadequate management from the healthcare system point of view; €524.60 per avoided inadequate management from the hospital point of view). In addition, from the healthcare system point of view, the GBS screening strategy using rapid PCR testing is more favourable than from the hospital point of view, as there is a probability of 16.4% that the ICER belongs to the south-east quadrant, in which the GBS screening strategy using rapid PCR testing at delivery dominates the GBS screening strategy using conventional culture at 34–38 weeks of gestation. This means that rapid PCR testing at delivery is both less costly and more effective than conventional culture at 34–38 weeks of gestation.

In the sensitivity analysis, five scenarios for the unit cost of PCR testing were taken into consideration, as described in the footnotes of Table 5. From the healthcare system point of view, the sensitivity analysis shows that in scenarios 1 and 2 there is a probability of 87.0 and 99.8%, respectively, that the true ICER belongs to the south-east quadrant of the CE plane (Figure 5), in which the GBS screening strategy using rapid PCR testing at delivery is both less costly and more effective than the GBS screening strategy using conventional culture at 34–38 weeks of gestation. Thus, in both cases, which are the most optimistic in terms of PCR cost, rapid PCR testing at delivery dominates conventional culture at 34–38 weeks of gestation. In scenarios 4 and 5, which are the most pessimistic, the upper bounds of the 95% confidence interval of the ICER are, respectively, €310 and €370 per inadequate management avoided (ICER point estimates of €89 and €114 per avoided inadequate management, respectively).

Table 5. Sensitivity analysis showing the impact of the variation in the cost of PCR on incremental cost-effectiveness ratios (ICERs) and associated 95% confidence intervals from the healthcare system point of view
Various cases for unit cost of PCRICER (€/avoided unadequate management)Lower bound of the 95% confidence regionUpper bound of the 95% confidence regionProbability that the true ICER be in the north-east quadrant of the CE plane (%)Probability that the true ICER be in the south-east quadrant of the CE plane (%)Conclusion
  1. a

    Case 1: ‘lowest bound case’, corresponds to hospital low cost hypothesis (€25).

  2. b

    Case 2: ‘Base case−20%’, corresponds to a decrease of 20% in the cost of the PCR test (€32.40).

  3. c

    Case 3: ‘Base case’, corresponds to the hypothesis taking into account the cost given by the database of the University-Hospital of Montpellier (€40.50).

  4. d

    Case 4: ‘Highest bound case’, corresponds to hospital high cost hypothesis (€46.00).

  5. e

    Case 5: ‘Base case + 20%’, corresponds to an increase of 20% in the cost of the PCR test (€48.60).

Case 1a–112–274–420.299.8PCR cost–effective
Case 2b–42–134441387PCR cost-effective
Case 3c36–3419083.216.8PCR acceptable
Case 4d8913310991PCR acceptable
Case 5e1143237099.90.001PCR acceptable
Figure 5.

Sensitivity analysis showing the impact of the variation in the cost of PCR on incremental cost-effectiveness ratios (ICER) and associated 95% confidence intervals (95% CIs) from the healthcare system point of view. The graph shows ICER relative to a variation in the cost of PCR test according to the five scenarios described in the footnote of Table 5 (numbered 1 to 5). The dashed line is relative to ICER point estimates. The solid line plots the values of the upper bound of the ICER confidence intervals.

Discussion

Main findings

This ‘real-life’ study followed current French clinical practice in terms of the prevention of GBS infection in neonates (including the use of a vaginal swab and not a combined vaginal and rectal swab, as recommended in the USA),[25] except that the result of the conventional culture at 34–38 weeks of gestation was not taken into consideration for deciding to start an antimicrobial treatment during delivery. It clearly shows that the strategy using PCR is much more efficient than that recommended by standard guidelines for all of the indicators tested,[4, 5, 25] including the ratio of pregnant women receiving an inadequate antimicrobial treatment (primary end point of the study), and several other parameters depicted in Table 1 (higher sensitivity, specificity, and positive predictive value, higher number of uninfected women not treated).

The prevalence of vaginal GBS during the last term of pregnancy and at delivery was 12.5%, a figure very close to that of previous studies (12.5 and 12.1% in the studies of El Helali et al.[17] and Goodman et al.[9], respectively). Our results also confirm that vaginal GBS is intermittent in a high proportion of women (12.3% in this study; 11.2% in El Helali et al.[17]; 10.8% in Alfa et al.[26]). This finding explains the inadequacy of a strategy relying on the preterm detection of vaginal GBS to predict the usefulness of an antimicrobial prophylactic treatment at delivery.

Moreover, the study validated the feasibility of the GenExpert test on a routine basis: 24 hours a day, 7 days a week. By its simplicity, and with the absence of the step for extracting nucleic acids, in particular, this PCR test can be performed by people who are not trained in molecular biology. However, the presence of mucus may inhibit the PCR reaction,[27] so the swab should be wiped before starting the test. Consequently, from our experience, the test requires some special training that restricts its use to healthcare workers familiar with biological testing.

Strengths and weaknesses

From the sampling step to the phone call giving the answer of the test to the clinician, the median response time was 110 minutes in our study, including 70 minutes for technical time on the GenExpert machine, which is in accordance with previously published results.[3, 18] The technical performance of rapid molecular tests have made considerable progress; indeed, in 2006, the duration of the detection of GBS by PCR assay was 18 hours.[13] Manufacturers must continue working to reduce the technical time of PCR assays in order to save enough time for administering the optimal dose of antimicrobial prophylactic treatment, thereby efficiently protecting the neonate.[28] Considering women with a positive GBS colonisation status, 16.7% of women received an incomplete dose of antibiotic because of the delay generated by the PCR test, which probably constitutes a limit of this strategy. However, even if it is recommended to administer antibiotics for at least 4 hours before delivery, colonisation data suggest that an antimicrobial treatment duration of at least 2 hours before delivery might confer significant anti-streptococcal protection.[29] Despite this limitation, the strategy using the GenExpert test was more efficient than the one based on GBS detection by culture at 34–38 weeks of gestation. Actually, even in the case of a PCR positive result being available only after delivery, or if the delivered antibiotic dose was inadequate, adapted treatment and surveillance could be started rapidly in the neonate.

The PCR technique was shown to exhibit a good specificity (94.9%, Table 2), close to that described by El Halali et al. with the same test.[17] The small proportion of ‘false positive’ results observed by PCR could correspond to very low bacterial loads or to the presence of non-culturable bacteria (some of these women were also found to carry vaginal GBS at 34–38 weeks of gestation). Atkins et al.[13] also described a few subjects with vaginal GBS detected by PCR and not by culture, in relation with the presence of non-viable organisms in the swab. The presence of antagonistic micro-organisms that inhibit GBS growth in culture has been also suggested.[30]

In this study, the sensitivity of the PCR test was relatively weak (66.7%, Table 2) in comparison with that of the 98.5% reported by El Halali et al. with the same test.[17] The nine culture-positive cases missed by the PCR assay could have corresponded either to true positive PCR results available only after delivery, or to weak positive culture. In the first case, it was possible to reinforce the surveillance of the newborn as soon as the positive result was available; the second situation is less preoccupying as low-level vaginal colonisation could reduce the risk of GBS transmission to the newborn.

Interpretation

The cost analysis was limited to the variable costs, with the fixed costs being assumed to be the same in each group. We thus did not consider the cost of labour of nurses supervising the presumably infected babies and of technicians and residents carrying out biological analyses, because we assumed that the labour cost was not a variable cost. Depending on the point of view adopted, the results of the cost analysis showed an ICER of €36 and €173 for the healthcare system and hospital, respectively. These ICERs seem relatively acceptable, especially if the costs of avoided complications or useless exposure to antibiotics are taken into account. Indeed, the extra costs must be put into balance against the several benefits linked to the improvement of patient management brought by PCR assay. First, there is a reduction in the time spent by nurses in administering useless antimicrobials and supervising uninfected neonates, together with an economy in antimicrobial consumption. Second, it could be expected that there would be a reduction in the rate of infection of neonates whose mothers were found to be negative for vaginal GBS at 34–38 weeks of gestation, but were positive for GBS at delivery. Even though the incidence of GBS neonatal infection dropped to 0.37 per 1000 births in 2004, the cost of a neonate infection with this bacterium can reach more than €19 000 (National Costs Scale 2007); moreover, for infants that survive the disease, the health and social care costs for the first 2 years of life have been estimated to be twice as high as those for unaffected infants because of possible disabilities.[31] Although these costs should bear upon the total cost, the present study lacked power to estimate the real incidence of infection in neonates. Third, the PCR strategy significantly reduces the number of useless antimicrobial treatments. Antimicrobial stewardship is a key component of a multifaceted approach to preventing the emergence of antimicrobial resistance: good antimicrobial stewardship involves selecting an appropriate drug and optimising its dose and duration to cure an infection, while minimising toxicity and conditions for selection of resistant bacterial strains.[32] Presently in France, as in America, all women in premature labour, i.e. before 37 weeks of gestation, are recommended to receive antibiotics if the GBS vaginal status is unknown. The detection of GBS by PCR at delivery in this population, which currently numbers over 10% of all deliveries, would dramatically reduce the number of women receiving antibiotics and, consequently, the number of fetuses and newborns exposed to antibiotics, with potentially long-term implications, notably in terms of allergy, asthma, and obesity. Although the S. agalactiae species remains sensitive to beta-lactams, the prevalence of the resistance to clindamycin and erythromycin has increased over recent years.[33]

To assess the robustness of our results, we performed a sensitivity analysis using five scenarios. Indeed, the real cost of the tests vary around the cost given by the database of the University-Hospital of Montpellier, according to the manufacturers' fees. In the next few years, with the generalisation of molecular tests, we could expect to observe a reduction of these costs, making the PCR assay even more cost-effective.

We chose to assess hospital and healthcare system perspectives. For the hospital, the results are informative for biologists and hospital decision makers, but such an innovative strategy cannot be restricted to a hospital, and involves political decisions from healthcare system decision makers. Using a before-and-after methodology, the work recently published by El Helali et al. demonstrates the cost-effectiveness of a strategy using the GenExpert test at delivery for a long period of time on a large cohort of subjects.[20]

Conclusions

The intrapartum detection of vaginal GBS by the GenExpert PCR assay was found to be effective to rationalise the use of antimicrobial drugs during labour. It was shown to be easy to implement in a hospital with biological facilities available 24 hours a day, 7 days a week. However, the change in the GBS screening strategy is beyond hospital policy. Cost-effectiveness studies like that of El Helali et al.[20] and this will help health authorities to decide whether molecular testing of GBS at delivery can be included with new recommendations.

Disclosure of interests

All authors declare that they have no conflicts of interest.

Contribution to authorship

Design of the study: AR, BP, and MNV. Design of the economic evaluation: FF. Data acquisition and management: EPJ. Analysis of data: CS, FF, and EPJ. Drafting the article: FF, EPJ., and BP. Critical review and approval of the final version of the article: all authors.

Details of ethics approval

This project was approved by the ethics committee of the University-Hospital of Saint-Etienne on 5 February 2009.

Funding

The study was funded by a grant from the Department of Clinical Research of the University-Hospital of Saint-Etienne.

Acknowledgements

The authors thank all the midwives, biological technicians, and residents in biology for their active participation in the study. All of the pregnant women who agreed to join this study are wholeheartedly acknowledged.

Reviewer's commentary on ‘Reduction of the use of antimicrobial drugs following the rapid detection at delivery of Streptococcus agalactiae in vagina by real-time PCR assay’

Two strategies exist to limit the acquisition of neonatal Group B streptococcal (GBS) infection. With the first complex strategy used both by Poncelet-Jasserand et al. in this BJOG issue and universally in developed countries, GBS identified in the maternal vagina at 35–37 weeks of gestation is treated with maternal antibiotics administered during labour. Although this strategy has very effectively reduced neonatal GBS sepsis from 2 to 0.3–0.4 cases per 1000 live births (Verani et al. MMWR 2010;59:1–31), both an immediate and a distant future problem exist. The immediate problem is to more accurately determine which baby is exposed to GBS at the time of labour. This first strategy is not useful for babies born before 35 weeks of gestation. Additionally, given the intermittent presence of GBS, the culture at 35–37 weeks of gestation both misses GBS that is subsequently acquired and misidentifies for treatment GBS that subsequently disappeared before labour. Further, the logistics are complicated to identify the 35–37-week test interval, to record results in medical records and to insure that the GBS-positive group is treated in labour. These logistical errors cause 15% of GBS-positive patients to not be treated in labour (Verani et al. MMWR 2010;59:1–31).

A rapid accurate polymerase chain reaction (PCR) test performed in labour would solve the problem of early gestation delivery before 35 weeks and the transient presence of GBS, and would simplify the logistics. The authors of this paper used a rapid PCR test in an attempt to identify GBS in early labour. They demonstrated a very acceptable time of 2 hours 20 minutes to perform the test, but this test had an unacceptably low sensitivity. Incubation for 18–24 hours in enrichment broth improves detection, but the time to develop an enriched sample limits its use in labour. Still, although not currently recommended, it is worth continuing to explore this technology because others report that the sensitivity of amplified GBS tests is >90%, even on samples not enriched (Verani et al. MMWR 2010;59:1–31). Additionally, transmission of vaginal GBS to the neonate appears much reduced (but not zero) at very low concentrations and future studies should examine the GBS concentration in PCR-negative, culture-positive cases to determine the actual risk of neonatal GBS acquisition at low GBS concentrations.

A distant future problem of this first strategy is antibiotic resistance. Up to 50% of women in labour now receive antibiotics for GBS, ruptured membranes, clinical chorioamnionitis and prophylaxis before caesarean section. This amount of antibiotic exposure of labouring women is a concern. Today, GBS remains exquisitely sensitive to most antibiotics, but after several decades of antibiotic exposure, we will start to find GBS resistance. Hence, an alternative strategy will eventually be needed to replace the present clumsy antibiotic strategy that will eventually become outmoded.

The second strategy to prevent neonatal GBS is to immunise the mother to develop high protective circulating levels of specific anti-GBS antibody. High anti-GBS antibody levels can protect against neonatal invasive GBS disease such as pneumonia and meningitis. However, sufficient stimulation of maternal antibody levels has proved agonisingly difficult. Still, more money and effort need to be placed toward this effort to eventually replace antibiotic treatment.

Disclosure of interests

I declare no conflicts of interest.

  • D Eschenbach

  • Department of Obstetrics and Gynecology, University of Washington, Seattle, WA, USA

Ancillary