PowerPoint Presentation - The HRB Centre for Primary Care Research

A systematic review of the diagnostic accuracy of signs and symptoms and
the validation of the Centor score in predicting group A β-haemolytic
streptococcal pharyngitis in adults in Primary Care
J Aalbers1,2, KK O’Brien1, WS Chan1, BD Dimitrov1, G Falk1 C Teljeur1 & T Fahey1
HRB Centre for Primary Care Research, RCSI1 & Radboud University, Nijmegen Medical Centre2
INTRODUCTION
Upper respiratory tract infections, including acute pharyngitis, are common in
general practice. Although the most common cause of pharyngitis is viral,
approximately 10% of incidences are bacterial, and are mostly accountable by
group A β-haemolytic streptococci (GABHS). Unlike other forms of pharyngitis,
GABHS pharyngitis requires treatment with antibiotics. Antibiotics prevent serious
complications (e.g. rheumatic fever), reduce duration of symptoms and spread of
disease. GABHS can be diagnosed with a throat swab, however, this test is
relatively expensive and the results can take days to come back, leading to
withholding of treatment or the prescription of unnecessary antibiotics. A number
of clinical prediction rules have been developed to distinguish streptococcal
pharyngitis from other types of pharyngitis using signs and symptoms. The most
widely recognised of these is the Centor score1.
Figure 1 shows ROC-curves for
each sign and symptom. The
curves are close together,
suggesting the ability of any
individual sign or symptom to
discriminate GABHS from non
GABHS patients is similar. ‘Any
exudates’ has the highest
accuracy. The symbols around the
ROC-curves indicate individual
studies and visually demonstrates
the high level of heterogeneity.
The signs and symptoms included in the Centor Score can be found in Table 1a.
One point is assigned for the presence of each sign or symptom and the
probability of GABHS is based on the patients score (Table 1b). The Centor rule is
recommended in a number of guidelines from north America; including the
‘American College of Physicians-American Society of Internal Medicine’ guidelines
and the ‘Centres for Disease Control and Prevention’ (CDC) guidelines.
No sign or symptom on its own
was powerful enough to rule in or
out a diagnosis of GABHS.
a
Symptoms
1 tonsillar exudates
2 tender anterior cervical
adenopathy
3 absence of cough
4 fever history (> 38.0C)
b
Points
1
1
Points
0
1
2
3
4
1
1
AUC =0.670
AUC =0.674
AUC =0.673
AUC =0.734
Figure 1. Summary ROC Curves for signs and
symptoms.
Validation of the Centor Score
Figure 2 & Table 3 show the results from validation studies of the Centor score.
There is no significant difference between predicted and observed values in any
of the Centor score categories, suggesting that overall the Centor score correctly
predicts the risk of GABHS pharyngitis. There is a tendency towards
underprediction in studies with a high prevalence of GABHS pharyngitis and
overprediction in studies with a low prevalence of GABHS pharyngitis.
Post-test Probability
2.5%
6.5%*
15.4%*
31.6%*
55.7%
Table 1. a) The Centor score, b) probability of GABHS based on score, from Centor et al 19811
*indicates average probabilities
No. of
Studies
No. of
Patients
Odds ratio and (95%
CI) or Range*
Isquared
0
1
2
3
4
10
10
12
12
13
423
682
675
456
215
0.57 (0.26-1.29)
0.67 (0.38-1.18)
0.55-1.47
0.84 (0.63-1.11)
1.24 (0.84-1.83)
0
36
55
0
0
Combined Centor
Score
0-1
2-3
4
11
13
13
1188
1243
215
0.66 (0.43-1.02)
0.62-1.22
1.24 (0.84-1.83)
25
63
0
Centor Score
AIM
The aim of this systematic review was twofold; to establish the discriminatory
power of individual signs and symptoms for ruling in or out a diagnosis of GABHS
pharyngitis in adults, and secondly to establish the validity of the Centor score.
METHOD
We searched PubMed (1966 to October 2008), EMBASE (1988 to October 2008),
Cochrane Library, Google Scholar and MEDION using a combination of terms and
filters. We included studies that assessed the diagnostic accuracy of signs and
symptoms and/or validated the Centor score using a throat swab as the gold
standard reference test. Only patients over the age of 15 years were included. For
the diagnostic accuracy of signs and symptoms, 2x2 tables were constructed and
sensitivities, specificities and SROC curves were calculated. To examine the validity
of the Centor score, the number of predicted GABHS patients (based on Centor’s
derivation study) was compared with the number of observed patients with GABHS
in each study. The quality of each study was assessed independently by two
researchers using a modified version of the QUADAS tool. Review Manager 5
(Cochran collaboration) and a bivariate random effects model were used for
analysis.
Table 3. Validation of the Centor Score for predicting GABHS pharyngitis
Centor Score 0-1
Centor Score 2-3
RESULTS
Search strategy identified 340 potentially relevant articles. After reading title,
abstract or full text, 34 of these articles fitted our inclusion criteria. 17 of the
articles included only adults. The authors of articles without the necessary data
and those who also included children were contacted. Data on the diagnostic test
accuracy of signs and symptoms was available in 18 studies. 13 studies also had
data on the Centor score. The majority of the studies were undertaken in a
primary care/family practice setting, two were carried out in an emergency
department. There was a good deal of variability in the prevalence of GABHS
between studies, ranging from 4.7% to 37.6%.
Centor Score 4
Diagnostic test accuracy of signs and symptoms
The signs and symptoms of ‘fever’ and ‘any exudate’ had a higher specificity than
sensitivity (Table 2), suggesting that they are more useful at ruling in a diagnosis
of GABHS when present. ‘Tender anterior cervical adenopathy’ and ‘absence of
cough’, on the other hand, have a higher sensitivity than specificity and are
therefore more useful at ruling out a diagnosis of GABHS when they are absent. It
should be noted that between study heterogeneity was high, however the bivariate
random effects model used in the analysis accounts for this.
Signs & Symptoms
Absence of
cough
Fever
Tender anterior
cervical
adenopathy
Any exudates
No. of
Studies
No. of
Patients
16
4027
Pooled
Sensitivity
0.73 (0.65-0.80)
0.50
(0.37-0.63)
1.45 (1.23-1.78)
0.55 (0.47-0.64)
18
13
4228
3528
0.48 (0.35-0.61)
0.67 (0.45-0.86)
0.72
0.61
(0.59-0.81)
(0.45-0.74)
1.68 (1.42-2.05)
1.70 (1.35-2.17)
0.73 (0.64-0.81)
0.54 (0.28-0.79)
18
4222
0.56 (0.42-0.68)
0.75
(0.66-0.83)
2.25 (1.80-2.88)
0.59 (0.46-0.72)
(95% CI)
Pooled
Specificity
(95% CI)
LR+
(95% CI)
LR-
Table 2. Pooled sensitivities and specificities, positive LR and negative LR calculated
using a bivariate random effects model
(95% CI)
Figure 2. Forest plots showing validation studies of the Centor score
CONCLUSIONS
No sign or symptom on its own was powerful enough to rule in or rule out
the diagnosis of GABHS pharyngitis.
However, our analysis suggests that the Centor score is a useful tool for
predicting the risk of GABHS pharyngitis and should be recommended in
UK & Irish guidelines, as it is in the US and Canada, so as to reduce
unnecessary prescribing of antibiotics.
REFERENCES
1Centor,
R.M., et al., The diagnosis of strep throat in adults in the emergency room. Med Decis Making, 1981.
1(3): p. 239-46 .