- Email: [email protected]
Infective endocarditis and antibiotic prophylaxis
Correspondence
suggests the need for additional analyses to establish a definite temporal association. BI reports personal fees from Abbott, Boehringer Ingelheim, Valtech, and Edwards Lifesciences, outside the submitted work. ST, FA, and ML declare no competing interests.
*Bernard Iung, Sarah Tubiana, François Alla, Marc Lavielle [email protected]
For the National Institute for Cardiovascular Outcomes Research see http://nicor4.nicor. org.uk
Cardiology Department, Assistance PubliqueHôpitaux de Paris, Hôpital Bichat, Paris, France (BI); Département Hospitalo-Universitaire Fibrosis Inflammation and Remodeling in cardiovascular, respiratory and renal diseases, Paris, France (BI); Infection Antimicrobials Modelisation Evolution, Institut National de la Santé et de la Recherche Médicale (INSERM) Unité 1137 (ST), Université Paris Diderot (BI), Sorbonne Paris Cité, Paris, France; Institut National de la Santé et de la Recherche Médicale (INSERM) Clinical Investigation Center 1425, Paris, France (ST); Maladies chroniques, santé perçue et processus d’adaptation, Equipe d’accueil 4360, Faculté de Médecine, Université de Lorraine, Vandoeuvre-lès-Nancy, France (FA); Institut National de la Santé et de la Recherche Médicale, (INSERM) Clinical Investigation Center, Centre Hospitalier Universitaire, Nancy, France (FA); and Institut National de la Recherche en Informatique et Automatique Saclay and University Paris-Sud, Orsay, France (ML) 1
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28.
Mark Dayer and colleagues1 report an increase in admissions to hospital with infective endocarditis after the UK National Institute for Health and Care Excellence (NICE) recommended co mplete cessation of antibiotic prophylaxis in 2008. We are concerned that the identified inflection point in admissions to hospital with infective endocarditis is a statistical artifact that can be anticipated when linear analytic methods are applied to non-linear data. On account of an ageing population, improved survival from cardiovascular disease, and an increasing number of invasive procedures each year, the number of patients at high risk of infective endocarditis is increasing in a non-linear manner. However, segmented linear regression models such as that used in this study assume that underlying time trends are linear.2 Application of these models 530
to curvilinear data might identify a spurious inflection point at the time of the prespecified intervention (an effect that increases with longer follow-up). Further supporting the argument that the linear model is inappropriate, the observed case fatality rate in the purported excess number of cases of infective endocarditis is substantially and inexplicably lower than historical values (4·2% vs 10–15%). Alternative statistical approaches include indexing admissions to hospital with infective endocarditis to all-cause admissions (which, incidentally, showed an apparent upward inflection in April, 2008),3 and investigation of regional differences in prescription practices and admissions to hospital with infective endocarditis. Adjustments that attenuate the magnitude or alter the timing of the reported association between publication of the NICE guidelines and subsequent increase in admissions to hospital with infective endocarditis would undermine the case for a causal association. As presented, this analysis is inadequate to alter the weight of evidence on which endocarditis prophylaxis guidelines are based. AFB is an author of the American Heart Association (AHA) Guidelines on the Prevention of Endocarditis, and is chair of the AHA Committee on Rheumatic Fever, Endocarditis and Kawasaki Disease. DSK declares no competing interests.
*Dhruv S Kazi, Ann F Bolger [email protected] Medicine (Cardiology) (DSK, AFB), Department of Epidemiology and Biostatistics (DSK), and Center for Healthcare Value (DSK), University of California San Francisco, CA, USA; and Division of Cardiology, San Francisco General Hospital, San Francisco, CA, USA (DSK, AFB) 1
2
3
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England, 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002; 27: 299–309. Health and Social Care Information Center. Hospital episode statistics. Admitted inpatient care: 2012–13. November, 2013. http://www. hscic.gov.uk/catalogue/PUB12566/hosp-episstat-admi-summ-rep-2012-13-rep.pdf (accessed July 15, 2015).
Mark Dayer and colleagues1 outline an increased incidence of infective endocarditis since cessation of antibiotic prophylaxis for dental procedures in the UK. However, as the authors clearly point out, they are unable to show a causal association. The increase might be ascribed to several factors, such as improved coding of infective endocarditis, which was incomplete before 2008. Since 2009, infective endocarditis has almost become a notifiable disease, with reporting of cases to the National Institute for Cardiovascular Outcomes Research (NICOR) by all UK congenital cardiac centres. At our centre, 42 patients with congenital heart disease have had infective endocarditis since 2008. In 24 (57%) of the patients, the detected organism was an oral pathogen, a similar proportion to that reported in the literature previously.2 Only two (8%) of these patients had a temporal, but not necessarily causal, relation to a dental procedure without antibiotic prophylaxis. Importantly, five (21%) of 24 patients had dental caries or abscesses needing remedial treatment at the time that infective endocarditis was diagnosed. In trying to identify small numbers of patients who might benefit from dental prophylaxis we might be missing another issue—promotion of dental hygiene. Oral microorganisms are a major cause of infective endocarditis, and improvement of overall hygiene could potentially prevent more cases than would antibiotic prophylaxis. We welcome the discussion that this Article has generated. A full assessment of the NICOR dataset might add clarity to this ongoing debate. TW is a member of the steering committee for the UK National Congenital Heart Disease Audit. Data cited here are reported to this audit. JW and ER declare no competing interests.
James Wong, Thomas Witter, *Eric Rosenthal [email protected] Department of Paediatric Cardiology, Evelina London Children’s Hospital, Guy’s and St Thomas’ National Health Service (NHS) Foundation Trust, London, UK (JW, TW, ER); and Department of Cardiovascular
www.thelancet.com Vol 386 August 8, 2015
Correspondence
Authors’ reply John Chambers and Jonathan Sandoe raise concerns that our data are limited by scarcity of microbiological information. We agree, and are pursuing this further. Data from Public Health England, the most robust data available, suggest a decline in staphylococcal bacteraemia and septicaemia between 2003 and 2012 and a corresponding increase in streptococcal bacteraemias (appendix). Furthermore, US studies have identified a concerning increase in streptococcal infective endocarditis since modification of the American College of Cardiology (ACC) and American Heart Association (AHA) guidelines in 2007.1 We reported the increasing number of device implants and valve replacements in our Article.2 However, UK data suggest a decrease in the number of intravenous drug users since 2004 (appendix).3 We acknowledge the importance of an accurate denominator and agree that an increasing number of patients might be at high risk of developing infective endocarditis.2 At present, however, we lack data quantifying the relative risk of infective endocarditis posed by different predisposing conditions and procedures, which makes full assessment of their effects on incidence difficult. Nonetheless, we were unable to identify a sudden change in volume of predisposing conditions and procedures that could explain the increased incidence of infective endocarditis.2 Restricted access to adequate dental care in England could underlie the increased incidence of infective endocarditis. Although the government report referred to by Chambers and Sandoe identified a decrease in dental attendances after a change www.thelancet.com Vol 386 August 8, 2015
Bernard Iung and colleagues, and Dhruv Kazi and Ann Bolger suggest alternative methods of statistical analysis, but ours was a prespecified approach recommended by Wagner and colleagues 4 as robust for studies like ours with a fairly short post-intervention period. As a simple check for non-linearity, we did a log transform of the incidence data and still identified a significant change in slope (p=0·015). Change-point or break-point analysis is complex, and different methods often yield different results. We used Hinkley’s5 change-point analysis, implemented by R’s change-point package, version 1.1.5, to test the robustness of our intervention point. We agree that other methods might highlight a different range of points, especially when applied retrospectively. Kazi and Bolger suggest that admissions to hospital with infective endocarditis should be indexed to all-cause admissions. We have undertaken this analysis (figure), and the increase in incidence of infective endocarditis after March, 2008, becomes larger and more statistically significant with this approach, with the number of extra cases per month increasing from 34·9 to 43·2 (p<0·0001). We respectfully disagree that our analysis is inadequate to alter the
See Online for appendix
30
25
20
15
10
5 Change in level –0·595, 95% CI –2·47 to 1·28, p=0·534; change in slope 0·097, 95% CI 0·05–0·15, p<0·0001 0
, Ju 200 l 0 D e , 20 c, 00 2 Ju 00 l 0 De , 20 c, 01 Ju 200 l 1 De , 20 c, 02 Ju 200 l, 2 Jan 200 , 3 Ju 200 l 4 De , 20 c, 04 2 Ju 00 l 4 De , 20 c, 05 Ju 200 l 5 D e , 20 c, 06 2 Ju 00 l, 6 Jan 200 , 7 Ju 200 l, 2 8 De 0 c, 08 Ju 200 l 8 De , 20 c, 09 2 Ju 00 l 9 De , 20 c, 10 2 Ju 010 l, Jan 201 , 1 Ju 201 l, 2 2 De 0 c, 12 20 12
2
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England, 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28. Li W, Somerville J. Infective endocarditis in the grown-up congenital heart (GUCH) population. Eur Heart J 1998; 19: 166–73.
Discharges (superspells) or deaths per 100 000 all-cause hospital admissions
1
of UK National Health Service (NHS) reimbursement mechanisms in 2006, subsequent data show that this reduction was only small and temporary.2 However, we agree that a study investigating the relation between infective endocarditis and dental procedures is long overdue. The data highlighted by Behnood Bikdeli and colleagues, which suggest a decrease in admissions to hospital with infective endocarditis, are concerning in view of our findings and the disparity in international guidelines. However, several interpretations are possible and we agree that these studies are surveillance rather than true incidence studies. Additionally, US Nationwide Inpatient Sample Data suggest that diagnoses of infective endocarditis are increasing and incidence of streptococcal infective endocarditis has significantly increased since the 2007 AHA guideline revision.1 We wholly agree that a randomised controlled trial would be of immense value if funding could be gained and ethical barriers could be overcome. We agree with Frédérique Gouriet and colleagues that several possible explanations exist for our data. However, although we make no claim of causation for the increase in infective endocarditis since 2008, we have been unable to identify any satisfactory alternative explanations.
Jan
Imaging, Division of Imaging Sciences and Biomedical Engineering, King’s College London, King’s Health Partners, St Thomas’ Hospital, London, UK (JW)
Figure: Discharges or deaths from infective endocarditis The grey dashed line shows the trend line for infective endocarditis incidence before March, 2008. The red dashed line shows the trend for infective endocarditis incidence after March, 2008.
531
suggests the need for additional analyses to establish a definite temporal association. BI reports personal fees from Abbott, Boehringer Ingelheim, Valtech, and Edwards Lifesciences, outside the submitted work. ST, FA, and ML declare no competing interests.
*Bernard Iung, Sarah Tubiana, François Alla, Marc Lavielle [email protected]
For the National Institute for Cardiovascular Outcomes Research see http://nicor4.nicor. org.uk
Cardiology Department, Assistance PubliqueHôpitaux de Paris, Hôpital Bichat, Paris, France (BI); Département Hospitalo-Universitaire Fibrosis Inflammation and Remodeling in cardiovascular, respiratory and renal diseases, Paris, France (BI); Infection Antimicrobials Modelisation Evolution, Institut National de la Santé et de la Recherche Médicale (INSERM) Unité 1137 (ST), Université Paris Diderot (BI), Sorbonne Paris Cité, Paris, France; Institut National de la Santé et de la Recherche Médicale (INSERM) Clinical Investigation Center 1425, Paris, France (ST); Maladies chroniques, santé perçue et processus d’adaptation, Equipe d’accueil 4360, Faculté de Médecine, Université de Lorraine, Vandoeuvre-lès-Nancy, France (FA); Institut National de la Santé et de la Recherche Médicale, (INSERM) Clinical Investigation Center, Centre Hospitalier Universitaire, Nancy, France (FA); and Institut National de la Recherche en Informatique et Automatique Saclay and University Paris-Sud, Orsay, France (ML) 1
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28.
Mark Dayer and colleagues1 report an increase in admissions to hospital with infective endocarditis after the UK National Institute for Health and Care Excellence (NICE) recommended co mplete cessation of antibiotic prophylaxis in 2008. We are concerned that the identified inflection point in admissions to hospital with infective endocarditis is a statistical artifact that can be anticipated when linear analytic methods are applied to non-linear data. On account of an ageing population, improved survival from cardiovascular disease, and an increasing number of invasive procedures each year, the number of patients at high risk of infective endocarditis is increasing in a non-linear manner. However, segmented linear regression models such as that used in this study assume that underlying time trends are linear.2 Application of these models 530
to curvilinear data might identify a spurious inflection point at the time of the prespecified intervention (an effect that increases with longer follow-up). Further supporting the argument that the linear model is inappropriate, the observed case fatality rate in the purported excess number of cases of infective endocarditis is substantially and inexplicably lower than historical values (4·2% vs 10–15%). Alternative statistical approaches include indexing admissions to hospital with infective endocarditis to all-cause admissions (which, incidentally, showed an apparent upward inflection in April, 2008),3 and investigation of regional differences in prescription practices and admissions to hospital with infective endocarditis. Adjustments that attenuate the magnitude or alter the timing of the reported association between publication of the NICE guidelines and subsequent increase in admissions to hospital with infective endocarditis would undermine the case for a causal association. As presented, this analysis is inadequate to alter the weight of evidence on which endocarditis prophylaxis guidelines are based. AFB is an author of the American Heart Association (AHA) Guidelines on the Prevention of Endocarditis, and is chair of the AHA Committee on Rheumatic Fever, Endocarditis and Kawasaki Disease. DSK declares no competing interests.
*Dhruv S Kazi, Ann F Bolger [email protected] Medicine (Cardiology) (DSK, AFB), Department of Epidemiology and Biostatistics (DSK), and Center for Healthcare Value (DSK), University of California San Francisco, CA, USA; and Division of Cardiology, San Francisco General Hospital, San Francisco, CA, USA (DSK, AFB) 1
2
3
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England, 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002; 27: 299–309. Health and Social Care Information Center. Hospital episode statistics. Admitted inpatient care: 2012–13. November, 2013. http://www. hscic.gov.uk/catalogue/PUB12566/hosp-episstat-admi-summ-rep-2012-13-rep.pdf (accessed July 15, 2015).
Mark Dayer and colleagues1 outline an increased incidence of infective endocarditis since cessation of antibiotic prophylaxis for dental procedures in the UK. However, as the authors clearly point out, they are unable to show a causal association. The increase might be ascribed to several factors, such as improved coding of infective endocarditis, which was incomplete before 2008. Since 2009, infective endocarditis has almost become a notifiable disease, with reporting of cases to the National Institute for Cardiovascular Outcomes Research (NICOR) by all UK congenital cardiac centres. At our centre, 42 patients with congenital heart disease have had infective endocarditis since 2008. In 24 (57%) of the patients, the detected organism was an oral pathogen, a similar proportion to that reported in the literature previously.2 Only two (8%) of these patients had a temporal, but not necessarily causal, relation to a dental procedure without antibiotic prophylaxis. Importantly, five (21%) of 24 patients had dental caries or abscesses needing remedial treatment at the time that infective endocarditis was diagnosed. In trying to identify small numbers of patients who might benefit from dental prophylaxis we might be missing another issue—promotion of dental hygiene. Oral microorganisms are a major cause of infective endocarditis, and improvement of overall hygiene could potentially prevent more cases than would antibiotic prophylaxis. We welcome the discussion that this Article has generated. A full assessment of the NICOR dataset might add clarity to this ongoing debate. TW is a member of the steering committee for the UK National Congenital Heart Disease Audit. Data cited here are reported to this audit. JW and ER declare no competing interests.
James Wong, Thomas Witter, *Eric Rosenthal [email protected] Department of Paediatric Cardiology, Evelina London Children’s Hospital, Guy’s and St Thomas’ National Health Service (NHS) Foundation Trust, London, UK (JW, TW, ER); and Department of Cardiovascular
www.thelancet.com Vol 386 August 8, 2015
Correspondence
Authors’ reply John Chambers and Jonathan Sandoe raise concerns that our data are limited by scarcity of microbiological information. We agree, and are pursuing this further. Data from Public Health England, the most robust data available, suggest a decline in staphylococcal bacteraemia and septicaemia between 2003 and 2012 and a corresponding increase in streptococcal bacteraemias (appendix). Furthermore, US studies have identified a concerning increase in streptococcal infective endocarditis since modification of the American College of Cardiology (ACC) and American Heart Association (AHA) guidelines in 2007.1 We reported the increasing number of device implants and valve replacements in our Article.2 However, UK data suggest a decrease in the number of intravenous drug users since 2004 (appendix).3 We acknowledge the importance of an accurate denominator and agree that an increasing number of patients might be at high risk of developing infective endocarditis.2 At present, however, we lack data quantifying the relative risk of infective endocarditis posed by different predisposing conditions and procedures, which makes full assessment of their effects on incidence difficult. Nonetheless, we were unable to identify a sudden change in volume of predisposing conditions and procedures that could explain the increased incidence of infective endocarditis.2 Restricted access to adequate dental care in England could underlie the increased incidence of infective endocarditis. Although the government report referred to by Chambers and Sandoe identified a decrease in dental attendances after a change www.thelancet.com Vol 386 August 8, 2015
Bernard Iung and colleagues, and Dhruv Kazi and Ann Bolger suggest alternative methods of statistical analysis, but ours was a prespecified approach recommended by Wagner and colleagues 4 as robust for studies like ours with a fairly short post-intervention period. As a simple check for non-linearity, we did a log transform of the incidence data and still identified a significant change in slope (p=0·015). Change-point or break-point analysis is complex, and different methods often yield different results. We used Hinkley’s5 change-point analysis, implemented by R’s change-point package, version 1.1.5, to test the robustness of our intervention point. We agree that other methods might highlight a different range of points, especially when applied retrospectively. Kazi and Bolger suggest that admissions to hospital with infective endocarditis should be indexed to all-cause admissions. We have undertaken this analysis (figure), and the increase in incidence of infective endocarditis after March, 2008, becomes larger and more statistically significant with this approach, with the number of extra cases per month increasing from 34·9 to 43·2 (p<0·0001). We respectfully disagree that our analysis is inadequate to alter the
See Online for appendix
30
25
20
15
10
5 Change in level –0·595, 95% CI –2·47 to 1·28, p=0·534; change in slope 0·097, 95% CI 0·05–0·15, p<0·0001 0
, Ju 200 l 0 D e , 20 c, 00 2 Ju 00 l 0 De , 20 c, 01 Ju 200 l 1 De , 20 c, 02 Ju 200 l, 2 Jan 200 , 3 Ju 200 l 4 De , 20 c, 04 2 Ju 00 l 4 De , 20 c, 05 Ju 200 l 5 D e , 20 c, 06 2 Ju 00 l, 6 Jan 200 , 7 Ju 200 l, 2 8 De 0 c, 08 Ju 200 l 8 De , 20 c, 09 2 Ju 00 l 9 De , 20 c, 10 2 Ju 010 l, Jan 201 , 1 Ju 201 l, 2 2 De 0 c, 12 20 12
2
Dayer MJ, Jones S, Prendergast B, Baddour LM, Lockhart PB, Thornhill MH. Incidence of infective endocarditis in England, 2000–13: a secular trend, interrupted time-series analysis. Lancet 2015; 385: 1219–28. Li W, Somerville J. Infective endocarditis in the grown-up congenital heart (GUCH) population. Eur Heart J 1998; 19: 166–73.
Discharges (superspells) or deaths per 100 000 all-cause hospital admissions
1
of UK National Health Service (NHS) reimbursement mechanisms in 2006, subsequent data show that this reduction was only small and temporary.2 However, we agree that a study investigating the relation between infective endocarditis and dental procedures is long overdue. The data highlighted by Behnood Bikdeli and colleagues, which suggest a decrease in admissions to hospital with infective endocarditis, are concerning in view of our findings and the disparity in international guidelines. However, several interpretations are possible and we agree that these studies are surveillance rather than true incidence studies. Additionally, US Nationwide Inpatient Sample Data suggest that diagnoses of infective endocarditis are increasing and incidence of streptococcal infective endocarditis has significantly increased since the 2007 AHA guideline revision.1 We wholly agree that a randomised controlled trial would be of immense value if funding could be gained and ethical barriers could be overcome. We agree with Frédérique Gouriet and colleagues that several possible explanations exist for our data. However, although we make no claim of causation for the increase in infective endocarditis since 2008, we have been unable to identify any satisfactory alternative explanations.
Jan
Imaging, Division of Imaging Sciences and Biomedical Engineering, King’s College London, King’s Health Partners, St Thomas’ Hospital, London, UK (JW)
Figure: Discharges or deaths from infective endocarditis The grey dashed line shows the trend line for infective endocarditis incidence before March, 2008. The red dashed line shows the trend for infective endocarditis incidence after March, 2008.
531