REVIEW ARTICLE
Correspondence to: L Li, Department of Clinical Pharmacy, Shanghai Tenth Peoples Hospital, Tongji University School of
Medicine, Shanghai, China. Tel: (+86) 216 630 2761. Fax: (+86) 216 630 7668. e-mail: jiangruihua119@yahoo.cn
Table 1
1383
Chotmongkol,
Kashyap, 20079
Lopez-Cortes, 199510
Corral, 200411
Kashyap, 200612
Mishra, 199513
Choi, 200214
Rohani, 199515
Ribera, 198716
Coovadia, 198617
Patients
n
177
153
180
62
281
35
182
119
205
78
Test results
Quality score
Assay method
Cut-off
IU/l
TP
FP
FN
TN
Giusti method
Giusti method
Kinetic determination
Spectrophotometric method
Giusti method
Giusti method
Giusti method
Giusti method
Giusti method
Giusti method
15.5
5
10
8.5
11.39
5
10
9
9
10
12
55
10
8
96
24
21
14
32
11
11
12
11
6
10
3
6
13
2
18
4
11
10
6
21
3
15
0
0
4
42
75
149
42
154
5
140
92
171
45
STARD QUADAS
9
14
12
15
13
11
15
13
9
13
7
9
7
10
10
9
11
8
8
11
IU = international unit; TP = true-positive; FP = false-positive; FN = false-negative; TN = true-negative; STARD = standards for reporting diagnostic accuracy;
QUADAS = quality assessment for studies of diagnostic accuracy.
disagreements were resolved by consensus. Data retrieved from the reports included participant characteristics, test methods, sensitivity and specificity data,
cut-off values, publication year and methodological
quality. The numbers of true-positive, false-positive,
false-negative and true-negative results are shown in
Table 1.
We assessed the methodological quality of the
studies using guidelines published by the standards
for reporting diagnostic accuracy (STARD) initiative
(maximum score 25) and the quality assessment for
studies of diagnostic accuracy (QUADAS) tool (maximum score 14).18,19 In addition, for each study the
following study design characteristics were also retrieved: 1) cross-sectional design (vs. case-control design), 2) consecutive or random sampling of patients,
3) blinded (single or double) interpretation of determination and reference standard results, and 4) prospective data collection. If no data on the above criteria were reported in the primary studies, we requested
the information from the authors. If the authors did
not respond to our letters, the unknown items were
treated as no.
Statistical analysis
The kappa () coefficient was calculated as a measure
of agreement between two reviewers. We used standard methods recommended for meta-analyses of diagnostic test evaluations.20 The following measures
of test accuracy for each study, e.g., sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR) and diagnostic odds ratio (DOR)
were computed using several statistical software programmes. The analysis was based on a summary receiver operating characteristic (SROC) curve.20,21 The
sensitivity and specificity for the single test threshold
identified for each study were used to plot an SROC
curve.21 A random effects model was used to calculate the average sensitivity, specificity and other
measures across studies.22 The term heterogeneity
when used in relation to meta-analyses refers to the
degree of variability. Fishers exact tests were used to
detect significant heterogeneity. To assess the effects of
RESULTS
After independent review, 16 publications dealing
with ADA concentrations for the diagnosis of TBM
were considered to be eligible for inclusion in the analysis.814,2530 Studies were excluded if ADA concentration was determined only in TBM patients,30 they
recruited <10 patients with confirmed TBM,26 or
they did not calculate the sensitivity or specificity.25,27,29
Some of the data might reappear in the publication.13,28
A final 10 studies817 involving 375 patients with TBM
and 989 non-TBM patients were available for analysis; the clinical characteristics of these studies, along
with their QUADAS scores, are outlined in Table 1.
Quality of reporting and study characteristics
The average internal agreement between the two reviewers for items in the quality checklist was 0.9. The
average sample size of the included studies was 136
(range 35281). In all 10 studies,817 the diagnoses of
all patients with TBM were made based on positive
smears or cultures of M. tuberculosis in the CSF and/
or on clinical symptoms. There were also increased
proteins, decreased glucose and good response to antituberculosis drugs in the CSF. Our initial data were
affected by the poor quality of reporting in the primary studies. To overcome this problem, we contacted
all authors of the 10 studies by airmail as well as
e-mail where e-mail addresses were available. Six
1384
Table 2
TB patients/
non-TB subjects
Reference
standard
Cross-sectional
design
Consecutive
or random
Blinded
design
Prospective
Chotmongkol, 20068
Kashyap, 20079
Lopez-Cortes, 199510
Corral, 200411
Kashyap, 200612
Mishra, 199513
Choi, 200214
Rohani, 199515
Ribera, 198716
Coovadia, 198617
15/53
66/87
20/160
14/48
117/164
27/84
36/146
14/105
32/173
15/63
Bac /clin
Bac /clin
Bac /clin
Bac
Bac /clin
Bac /clin
Bac /clin
Bac /clin
Bac
Bac /clin
Yes
Yes
Yes
No
Yes
No
Yes
Unknown
Unknown
Unknown
Yes
Yes
Yes
Yes
Yes
Unknown
Yes
Yes
Yes
Unknown
No
No
Yes
No
No
No
No
Unknown
No
Unknown
Yes
Yes
Yes
No
Yes
Yes
No
Yes
Yes
Yes
Figure 1 Forest plot of estimate of sensitivity (A) and specicity (B) for adenosine deaminase assays in the diagnosis of TBM.
Circles = point estimates from each study; error bars = 95%CIs.
Pooled estimate of sensitivity for adenosine deaminase assays
was 0.79 (95%CI 0.750.83); pooled estimate of specicity
was 0.91 (95% CI 0.890.93). TBM = tuberculous meningitis;
CI = condence interval.
1385
Table 3 Weighted meta-regression of the effects of methodological quality and study design
on diagnostic precision of cerebrospinal adenosine deaminase in 10 assays
Covariates
STARD 13
QUADAS 10
Cross-sectional design
Consecutive or random
Blinded
Prospective
Studies
n
Coefcient
RDOR
P value
6
4
5
8
2
8
0.168
0.137
0.427
2.243
0.411
0.231
0.84 (0.441.63)
0.87 (0.391.97)
1.53 (0.337.10)
9.42 (0.61144.42)
0.66 (0.133.39)
1.26 (0.236.87)
0.5628
0.7041
0.5316
0.0932
0.5702
0.7565
RDOR = relative diagnostic odds ratio; STARD = standards for reporting diagnostic accuracy; QUADAS = quality
assessment for studies of diagnostic accuracy.
DISCUSSION
Figure 3 Funnel graph for the assessment of potential publication bias in adenosine deaminase assays. The funnel graph
plots the log of the OR against the SE of the log of the OR (an
indicator of sample size). Circles = each study in the metaanalysis. The result of the Egger test for publication bias was not
signicant (P = 0.171). OR = odds ratio; SE = standard error.
1386
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
1387
RSUM
O B J E C T I F : Dterminer la prcision des mesures de ladenosine deaminase (ADA) dans le diagnostic de la mningite tuberculeuse (TBM).
S C H M A : Aprs une revue systmatique des tudes en
langue anglaise, la sensibilit, la spcificit et la prcision
des concentrations de lADA pour le diagnostic dans le
liquide crbrospinal (CSF) ont fait lobjet dune valuation par les modles deffets alatoires. On a utilis les
courbes sommaires dopration du receveur pour rsumer les performances globales du test.
R S U LTAT S : Dix tudes ont rpondu nos critres din-
RESUMEN
O B J E T I V O : Determinar la exactitud de las mediciones
de desaminasa de adenosina (ADA) en el diagnstico de
la meningitis tuberculosa (TBM).
M T O D O : Despus de un anlisis sistemtico de las
publicaciones cientficas en ingls, se evalu la sensibilidad, la especificidad y la exactitud de las concentraciones del ADA del liquido cefalo-rachidiano (CSF) en el
diagnstico de la TBM mediante modelos de efectos
aleatorios. La eficacia global de la prueba se valid con
curvas de resumen de rendimiento diagnstico.
R E S U LTA D O S : Diez estudios cumplieron con los crite-