Rapid Responses to:

RESEARCH: Tim E Colbourn, Christian Asseburg, Laura Bojke, Zoe Philips, Nicky J Welton, Karl Claxton, A E Ades, and Ruth E Gilbert Preventive strategies for group B streptococcal and other bacterial infections in early infancy: cost effectiveness and value of information analyses BMJ 2007; 0: bmj.39325.681806.ADv2 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

The Dangers of Misinterpretation of Economic Analysis

Jude N Chukwuma, Chisolum I. Chukwuma, General Practitioner   (30 September 2007)

A misleading decision-analytic model

Peter Brocklehurst, Sara Kenyon, Alison Bedford-Russell, Malcolm Chiswick, Julia Drown, Rhona Hughes, Neil Marlowe, Mary Newburn, Stavros Petrou, Maria Quigley, Liz Schroeder, Ben Stenson, David Taylor, Joanne Woodward.   (20 October 2007)

Authors' response

Ruth E Gilbert, Tony Ades, Tim Colbourn, Karl Claxton   (29 October 2007)

The Dangers of Misinterpretation of Economic Analysis 30 September 2007
Jude N Chukwuma, Honorary Clinical Research Fellow School of Medicine, Swansea University, Chisolum I. Chukwuma, General Practitioner Send response to journal: Re: The Dangers of Misinterpretation of Economic Analysis	The findings by Colbourn et al(1) are very important and most welcome. However, it does bring to mind the dangers of political translations, interpretations and misinterpretations of such robust economic analyses. It will be wrong for politicians to harbour any notion that the National Health Service (NHS) is primarily a money making or money saving machine; the quality of life and associated productivity of a healthy population should always be borne in mind. Reference: 1. Colbourn TE, Asseburg C, Bojke L, Philips Z, Welton NJ, Claxton K, Ades AE, Gilbert RE. Preventive strategies for group B streptococcal and other bacterial infections in early infancy: cost effectiveness and value of information analyses. BMJ 2007; 335: 655-658 Competing interests: None declared
A misleading decision-analytic model 20 October 2007
Peter Brocklehurst, Professor of Perinatal Epidemiology National Perinatal Epidemiology Unit, University of Oxford Old Road Campus, Oxford OX3 7LF, Sara Kenyon, Alison Bedford-Russell, Malcolm Chiswick, Julia Drown, Rhona Hughes, Neil Marlowe, Mary Newburn, Stavros Petrou, Maria Quigley, Liz Schroeder, Ben Stenson, David Taylor, Joanne Woodward. Send response to journal: Re: A misleading decision-analytic model	Dear Sir We were interested to read the article by Colbourn and colleagues about the cost-effectiveness of screening for GBS infection in pregnancy to prevent neonatal sepsis. The authors conclusions suggest that current practice in the UK is not cost-effective. However, as with all complex models it is important to review the evidence that was used in the construction of the model and some of the assumptions made before deciding whether the results of this study are valid. We have identified a substantial number of assumptions which may impact on the robustness of the models findings. Those that are most serious include: All the “cost-effective” strategies include offering intrapartum antibiotics to all women presenting in preterm labour. The Cochrane review of antibiotics to prevent preterm birth for those women who currently present in preterm labour with intact membranes includes a number of randomised controlled trials using a variety of different antibiotic regimens (1). None of these are specifically about preventing mother to child transmission of GBS, but when the trials are combined in meta- analysis there is no evidence of benefit and a suggestion of an increased risk of neonatal mortality (Peto Odds Ratio 1.50, 95%CI 1.00 to 2.25). The Cochrane review authors therefore recommend that antibiotics are not used routinely for this group of women. Although none of these trials used benzyl penicillin, there are concerns that widespread use of antibiotics in this group of women may be detrimental to the fetus. The decision- analytic model did not include this important information. In addition, all the “cost-effective” strategies include giving antibiotic prophylaxis to all women undergoing elective caesarean section at term. The evidence base for including this group came from a national UK surveillance study where there was a single case of early-onset neonatal GBS sepsis in a woman who had an elective caesarean section (2). On detailed investigation of this report, the woman was found to have ruptured membranes prior to her planned caesarean section and was therefore misclassified as an elective caesarean section with intact membranes at term. As a result the number of women requiring antibiotics in all of the cost-effective screening strategies can be decreased by approximately 8-10%. In their model, the authors did not accept the findings of the Cochrane review of randomised controlled trials of intrapartum antibiotics for the prevention of mother to child transmission of GBS (3) and constructed a new meta-analysis, using Bayesian methods, and using different selection criteria to include a slightly different population of trials. The Cochrane review suggests that antibiotics are 80% to 90% effective at preventing infant colonisation and infant sepsis. The meta- analysis in the decision-analytic model suggests the effectiveness is over 97%. The authors have not considered the in-vitro evidence that intrapartum antibiotics are more likely to result in culture-negative sepsis (4). The presence of antibiotic in the infant’s bloodstream will prevent the culture of the organism, even though the baby may have clinical manifestations of infection. Culture negative sepsis makes up approximately 40% of all cases of neonatal sepsis. As all the existing randomised controlled trials have used positive cultures to assess efficacy, they will all tend to overestimate the efficacy of antibiotics in preventing neonatal sepsis. If a more ‘realistic’ estimate of the effectiveness of this treatment was included all the strategies would become more costly and less effective. Of interest is that if the apparently “cost-effective” strategies in this model do not include treating all women in preterm labour and do not include treating all women undergoing planned elective caesarean section, they resemble current UK practice (5). Finally, the proposed trial referred to by the authors was developed over 2 years by a group of 26 experts in the clinical management and epidemiology of perinatal GBS infection as well as clinical trialists, microbiologists and health economists (www.npeu.ox.ac.uk/gbstrial). After very careful review of the detailed modelling work undertaken by Colburn and colleagues, the trial protocol development group concluded that the serious flaws in the model, some of which are referred to above, meant that the findings were unreliable and that evidence of relative cost- effectiveness could not be used in designing subsequent research. The proposed trial, therefore, aims to compare current clinical practice with the screening strategy used in the USA and many other countries where observational data suggest this strategy is cost-effective. (1) King J, Flenady V. Prophylactic antibiotics for inhibiting preterm labour with intact membranes. Cochrane Database of Systematic Reviews 2002, Issue 4. Art. No.: CD000246. DOI: 10.1002/14651858.CD000246. (2) Heath PT, Balfour G, Weisnor AM, Efstratiou A, Lamagni TL, Tighe H, O’Connell LAF, Cafferkey M, Verlander NQ, Nicoll A, McCartney C on behalf of PHLS GBS Working Group. Group B streptococcal disease in UK and Irish Infants younger than 90 days. Lancet 2004, 363; 292-294. (3) Smaill F. Intrapartum antibiotics for Group B Streptococcal colonisation. In.Cochrane Library Issue 1. Oxford Update software 1996. (4) Hsu K, Pelton S, Shapiro D. Detection of group B streptococcal bacteremia in simulated intrapartum antimicrobial prophylaxis. Diagn Microbiol Infec Dis 2003;45:23-27. (5) RCOG 2003. Prevention of early onset neonatal group B streptococcal disease. Guideline Number 36, Available: http/www.rcog.org.uk. Competing interests: None declared
Authors' response 29 October 2007
Ruth E Gilbert, Professor of Clinical Epidemiology Centre for Paediatric Epidemiology and Biostatistics, UCL Institute of Child Health, London WC1N 1EH, Tony Ades, Tim Colbourn, Karl Claxton Send response to journal: Re: Authors' response	Authors’ response The group of experts led by Brocklehurst worked in parallel with our study as both groups were funded by UK Health Technology Assessment Programme. Brocklehurst et al held a series of meetings and drew on expert opinion to develop a trial protocol. Our study was based on the most comprehensive synthesis to-date of the research evidence, using state of the art methods to integrate all relevant data.1 Having met the group of experts led by Brocklehurst on several occasions and shared our results, we have previously discussed the criticisms raised in their letter and addressed them in the full report (chapters 3, 5 and 9).2 Our responses are given in brief below. In summary, none of the criticisms alter our main findings. The review by King and Flenady provides no clear evidence of adverse effects of penicillin for women in preterm labour with intact membranes.3 Firstly, the most relevant results are those for perinatal mortality, which were very close to unity. Focusing on neonatal death can be misleading as antibiotics may postpone stillbirth until the neonatal period. Secondly, any adverse effect on neonatal death appears to be confined to interventions involving combinations of antibiotics. Thirdly, evidence is lacking as to whether neonatal death was more likely to occur after bacteraemia in babies of treated compared with untreated women. The last point could be addressed by secondary analyses of the ORACLE dataset, a large trial coordinated by one of the experts in the group (SK), which contributed almost 90% of the women included in the review. Women delivering by elective caesarean section at term were estimated to have a very low, but not zero, risk of adverse outcome (Table 14, page 26).2 We recommended that such women should not be treated unless they have a positive test result (chapters 7 and 9). We did not accept the meta-analysis reported by Smaill et al as the Peto method is known to be biased when effect sizes are large (chapter 5, page 42).4 We have shown (submitted for publication, draft available from authors) that all the previous meta-analyses,5;6 including the one by Smaill,7 give results that are very similar to ours when statistical methods are used, which are valid for large effect sizes and trials with zero cell counts, for example the Mantel Haenzel and Exact methods. Brocklehurst et al are correct that we did not consider a commercially funded study showing that in vitro, resin impregnated culture medium was more sensitive for detecting group B streptococci than standard media.(4) We have yet to see clinical evidence to support the hypothesis, put forward by Brocklehurst, that failure to take account of culture negative sepsis leads to an overestimate of the treatment effect. However, we did use a less extreme treatment effect in the model, in the form of oral treatment. The cost effectiveness results were similar (chapter 6 and page 191), and far exceeded any benefits of prenatal screening for group B streptococcal colonisation in women with preterm or high risk term deliveries. Reference List (1) Briggs AH. New methods of analysing cost effectiveness. BMJ 2007; 335:622-623. (2) Colbourn T, Asseburg C, Bojke L, Phillips Z, Claxton K, Ades AE et al. Prenatal screening and treatment strategies to prevent group b streptococcal and other bacterial infections in early infancy: cost effectiveness and expected value of information analyses. Health Technol Assess 2007; 11(29). (3) King J, Flenady V. Prophylactic antibiotics for inhibiting preterm labour with intact membranes. Cochrane Database Syst Rev 2002;(4):CD000246. (4) Greenland S, Salvan A. Bias in the one-step method for pooling study results. Stat Med 1990; 9:247-252. (5) Allen UD, Navas L, King SM. Effectiveness of intrapartum penicillin prophylaxis in preventing early-onset group B streptococcal infection: results of a meta-analysis. CMAJ 1993; 149(11):1659-1665. (6) Benitz WE, Gould JB, Druzin ML. Antimicrobial prevention of early-onset group B streptococcal sepsis: estimates of risk reduction based on a critical literature review. Pediatrics 1999; 103(6):e78. (7) Smaill F. Intrapartum antibiotics for group B streptococcal colonisation. Cochrane Database Syst Rev 2000;(2):CD000115. Competing interests: None declared