1. Diagnosing Clostridium difficile. Tim Planche St. George’s Healthcare NHS Trust St. George’s, University of London

2. Laboratory CDI diagnosis change or confusion or both? There is no doubt that the laboratory diagnosis of CDI is in a state of flux For 10-15 years the laboratory diagnosis of CDI has relied mainly on the use of enzyme immunoassays (EIAs) to detect the presence of the major C. difficile toxin(s)

3. Problems with incorrect results False negatives Miss some cases Poor isolation Possible non/sub- optimal therapy Unnecessary investigation for other conditions False positives Missing real cause of disease Unnecessary therapy Unnecessary stopping of antibiotic therapy Cohorting with real cases of C. diff/MRSA/noro

4. Natural history of CDI Detection of toxigenic C. difficile 1-2% of healthy population 10-15% of non-diarhoeal hospital admissions 15-30 % of long stay non-diarrhoeal patients

5. Example of C. difficle Substance detectedTesting method C. difficile toxin Cell-cytotoxicity assay Immunoassays (EIA & membrane) C. Difficile spores Culture Antigen (GDH) detection Toxogenic C. difficile spore Cytotoxigenic culture PCR

6. Reference/Gold standards Define True cases of a disease Used for comparison of any newer test I often use the false positivity rate Also can use diagnostic likelihood ratios

7. Problems with reference standards BUT Problems if newer standard is better Problems if more than one standard - Need for clinical validation

8. Reference tests for C. difficle Substance detectedTesting method C. difficile toxin Cell-cytotoxicity assay Immunoassays (EIA & membrane) C. Difficile spores Culture Antigen (GDH) detection Toxogenic C. difficile spore Cytotoxigenic culture PCR Cell-cytotoxicity assay – detects toxin directly in stool – uses cytopathic effects on cultured cells Cytotoxigenic culture – cultures after alcohol shock – phenotypic ID and then detection of toxin

9. Example of C. difficile?

10. Clinical significance of reference assays 2 studies from 1986 – not much else Gerding et al 1986 – 149 cases either 109 CCTA +ve 40 only TC +ve – 148 non-diarrhoeal controls Controls no different from 40 TC +ve only (except fever) CCTA +ves had higher WCC, fever, longer hospital stay – 96 cases had colonoscopy 35/68 CCTA +ve had PMC 2/27 TC only +ve had PMC

11. Clinical significance of reference assays Lashner et al 1986 Patients with diarrhoea – 29 untreated TC+ve/CCTA –ve cases – 22 cases had outcome data – 17 fully recovered – 1 had colectomy for IBD – 4 died - cause not recorded – 13 cases had diagnostic studies – 6 IBD 7 normal

12. test performance in general Sens 50% Spec 100% Sens 100% Spec 75% Sens 80% Spec 98% ROC Curve

13. Systematic Review of Studies - results

14. How do we know how accurate our numbers are? Roll 60 dice – obviously will not always roll 10 sixes exactly Equally any study only estimates the proportion

15. The PPV and NPV vary

16. State of testing in the UK over the last decade

17. St. George’s number of monthly tests

18. positivity ratenumber of (prevalence) tests 20086.45% 4208 20094.47% 4241 Of 108 labs using single EIA, 39 (36%) using a test with estimated * PPV <50% *Planche T et al. Lancet Infect Dis 2008. Laboratory diagnosis of CDI FOI survey April, 2010 (n=168/170 trusts, England) Goldenberg SD, French GL, J Hosp Infect 2011.

19. Laboratory diagnosis of CDI FOI survey April, 2010 (n=168/170 trusts, England) Survey carried out 1 year after DH/HPA warning & guidance – 70% continue to use EIA as single test – 21% 2-stage test – 3.6% cytotoxin test – 0.6% (n=1) PCR Goldenberg SD, French GL, J Hosp Infect 2011.

20. Calculated PPV – example of C. difficile Calculated PPV Prevalence of C. diff

21. CEP report CEP08054 – Wilcox et al 2009 Conclusion The poor PPVs of toxin detection kits, especially in the context of widespread testing, raises doubts about their appropriateness when used as single tests for the laboratory detection of C. difficile toxins. http://www.pasa.nhs.uk/pasa/Doc.aspx?Path=%5bMN%5d%5bS P%5d/NHSprocurement/CEP/CEP08054.pdf

22. Solutions 1) better test 2) combine tests

23. Other Future Possibilities Look at newer diagnostics – such as the stool proteome Consider using markers of gut inflammation such as calprotectin or lactoferrin

24. Disadvantages of NAATs for CD toxin genes NAATs do not detect the presence of faecal toxin. inpatients typically have C. difficile toxigenic culture positive rates of 10-20%. Colonization by C. difficile is protective against the development of CDI particularly when accompanied by an antitoxin antibody response.

25. Disadvantages of NAATs for CD toxin genes (ii) Diarrhoea is a common symptom, especially in elderly inpatients, many of whom will have received antibiotics or laxatives, or have been exposed to other pathogens (notably norovirus) Variable potential therefore to detect toxigenic C. difficile as an innocent (possibly protective) bystander.

26. General Ways to combine tests Unaltered Cut-offs Modified Cut-offs One after the other Simultaneous

27. Tests in series But number of repeats varies

28. 12 months, 4 hospitals 4-5 months Assess 1-2 two stage algorithms Optimise cut-offs and protocols compared to reference methods (n = 7000) Stage 1 training set Optimising the diagnosis of CDI - study plan Stage 2 testing set Final assessment of optimised algorithm vs reference methods (n = 5500) Faecal samples taken for routine diagnosis tested by both reference and study methods Faecal samples taken for routine diagnosis tested by both reference and study methods AnalysisSet-up and preparation Selection of preferred algorithm(s) Finished collecting clinical data 31/10/11 June 2010

29. Preliminary data (n~7000) reference standard = cytotoxigenic culture single assays -manufacturers' cut-offstwo-stage assays - manufacturers' cut-offs GDHEIA tox 1NAATEIA tox 2GDH NAAT EIA tox 1NAATEIA tox 2 EIA tox 1 93.6% (91.2- 95.5%) 46.9% (42.6- 51.2%) 93.0% (90.6- 95.0%) 56.6% (52.4- 60.8%) 55.0% (50.7- 59.2%) 89.8% (87.0- 92.2%) 55.4% (51.1- 59.5%) 56.4% (52.2- 60.6%) 46.7% (42.4- 50.1%) 94.5% (93.9- 95.0%) 99.2% (99.0- 99.4%) 97.1% (96.7- 97.5%) 98.7% (98.4- 99.0%) 99.6% (99.4- 99.7%) 98.1% (97.7- 98.4%) 99.5% (99.3- 99.6%) 99.6% (99.4- 99.7%) single assays -manufacturers' cut-offstwo-stage assays - manufacturers' cut-offs GDHEIA tox 1NAATEIA tox 2GDH NAAT EIA tox 1NAATEIA tox 2 EIA tox 1 92.1% (91.4- 92.8%) 69.2% (64.3- 73.8%) 94.9% 94.3- 95.4%) 82.3% (78.1- 85.9%) 67.4% (62.4- 72.1%) 94.6% (91.9- 96.6%) 80.4 (76.2- 84.3%) 82.0% (77.8- 85.7%) 68.9% (64.0- 73.6%) 95.9 (93.4- 97.6%) 99.4% (99.2- 99.6%) 96.9% (94.7- 98.4%) 98.8% (98.5- 99.1%) 99.7% (99.5- 99.8%) 95.6% (95.5- 96.5%) 99.6% (99.4- 99.7%) 99.6% (99.4- 99.8%) 99.7% (99.6- 99.8%) reference standard = cytotoxin test

30. Further work Analysis of the testing set What are the questions Need for the newer clinical validation

31. Thank you Leeds Kerrie Eastwood Mark Wilcox UCLH Nandini Shetty Mike Wren Pietro Coen Oxford Derrick Crook Sarah Oakley Lorraine Clark John Finney St. George’s Cassie Pope Irene Monahan