By Jolien Aalbers University of Nijmegen Streptococcal Pharyngitis

By Jolien Aalbers University of Nijmegen Streptococcal Pharyngitis: A Systematic Review of the Predictive Value of Signs and Symptoms and the External Validation of the Centor Score CPR Divisional Meeting 15 th January 2009

Outline • Background • Methods 1. Diagnostic accuracy of signs and symptoms 2. Validation of the Centor Score • Results 1. Diagnostic accuracy of signs and symptoms 2. Validation of the Centor Score

Background • Group A β-haemolytic Streptococcal (GABHS) pharyngitis can lead to serious complications • GABHS infection is the only indication for treatment with antibiotics • Antibiotic resistance is an important public health issue • A prediction rule to reduce unnecessary antibiotic prescriptions would help to decrease antibiotic resistance

Clinical prediction rule “A tool that quantifies the contribution of symptoms, clinical signs and available diagnostic tests to stratify patients according to the probability of having a target disorder” “The outcome can have a diagnostic, prognostic or therapeutic value”

Levels of evidence for the development of a CPR

The Centor Score 4 items: 1. Tonsillar exudate 2. Tender cervical anterior adenopathy 3. History of fever (or >38. 0 C) 4. Absence of cough Cumulative score (0 -4 points): 0. 2. 5% (post-test probabilities) 1. 6. 5% (6. 0 -6. 9) 2. 15. 4% (14. 1 -16. 6) 3. 31. 6% (30. 1 -34. 1) 4. 55. 7%

Other diagnostic tests Throat swab: “gold” standard • Takes 24 -48 hours for results • Expensive • A proportion of patients with positive cultures are carriers Rapid Antigen Detection Test (RADT): • Not as sensitive as the throat swab • Possible lack of cost-effectiveness

Study methods (1) Search strategy: • Pub. Med, EMBASE, Cochrane database, Google Scholar and MEDION • Hand-checking references of filtered papers Inclusion and exclusion-criteria: • Population: participants were recruited upon 1 st presentation from an ambulatory care setting with sore throat as their main presenting complaint, and had to be at least 15 years of age or older

Study methods (2) • Study design: the studies had to assess either the diagnostic accuracy of signs and symptoms and/or apply the Centor score • Reference standard: a throat swab • Retrospective and prospective studies were included • No criteria were set for the size of the population Quality assessment: modified QUADAS tool (QUality Assessment of Diagnostic Accuracy Studies)

Methods (3): data extraction and analysis of diagnostic accuracy of signs and symptoms Signs and symptoms: Exudate Absence of cough Fever (>38. 0 C) Tender Adenopathy • Data extraction from: 2 x 2 tables • Analysis with: likelihood ratios, ROC-curves, probability nomogram • I²-index for heterogeneity • Where heterogeneity is acceptable, likelihood ratios can be pooled

Methods (4): data extraction and analysis of validation of the Centor score • Predicted versus observed • Forest plots – using Random effect model • I²-index for heterogeneity • Prevalence correction

Results (1) • 340 potentially relevant articles • 16 studies included - 16 studies: signs and symptoms: 3371 patients - 11 studies: validation of the Centor score: 1603 patients • Large variability in prevalence: 4. 7% - 36. 8%

Positive likelihood ratios Sign of “any exudates” Pooled Positive LR = 1. 95 (1. 63 – 2. 32) I² = 74. 2% (heterogeneity)

Negative likelihood ratios Sign of “Absence of cough” Pooled Negative LR = 0. 59 (0. 50 – 0. 69) I² = 22. 7% (heterogeneity)

Results (2): Diagnostic accuracy of signs and symptoms • Very heterogeneous results • No powerful likelihood ratios Conclusion: no sign or symptom on its own is powerful enough to rule in or rule out the diagnosis of streptococcal pharyngitis

Results (3): Validation of the Centor score The Centor score four items: 1. Tonsillar exudate 2. Tender cervical anterior adenopathy 3. History of fever (or >38. 0 C) 4. Absence of cough Each item gets 1 point. Score: 0 -4 points

Forest plot for Centor score 0 -1 I² = 5% of heterogeneity Z = 2. 77, P = 0. 006

Forest plot for Centor score 2 -3 I² = 62% of heterogeneity Z = 1. 15, P = 0. 25

Forest plot for Centor score 4 I² = 7% of heterogeneity Z = 1. 33, P = 0. 18

Prevalence correction Centor Score 1 Adjusted for pooled prevalence of 21. 9%

Results (4): Validation of the Centor score Conclusions: • Heterogeneity is acceptable • Underprediction for Centor score 0 -1 • No significant difference between predicted and observed values in Centor score 2 -4 • Underprediction in studies with a higher prevalence and overprediction in studies with a lower prevalence, compared with Centors’ prevalence • Correction with a pooled prevalence is possible