Page 8 - Cardiac testing for coronary artery disease in potential kidney transplant recipients

HSROC model results were used to create plots of estimated sum- definitions of an abnormal test were present, only data that had
mary curves, summary points and confidence regions, superim- been measured using the same definitions were combined.
posed on study-specific estimates of sensitivity and specificity.
We provided summary measures of diagnostic accuracy for:
Sensitivity analyses
1. all studies regardless of CAD threshold on coronary
angiography Where differences were present across studies, we controlled for
heterogeneity by conducting sensitivity analyses. In particular, we
2. studies that reported ≥ 70% stenosis threshold for
diagnosis of significant CAD on coronary angiography. investigated diagnostic accuracy in studies that:
1. aimed to provide both index tests and reference tests to
Pairwise comparisons of test performance among alternative index
tests were performed using data from studies where comparisons their study population (studies that avoided verification bias)
2. applied a threshold of diagnosis of severe CAD of ≥ 70%
between tests were made in the same study population (direct com-
parison) or in different study populations (indirect comparison). stenosis on coronary angiography
3. consisted entirely of asymptomatic individuals (studies that
A covariate of test type was included in the modelling to assess if
the SROC curves for tests differed in shape, or overall accuracy. excluded patients who had either symptoms of cardiac disease or
a history of ischaemic heart disease).
When comparing the relative performance of two index tests, we
initially assumed equal variances for random effects for the tests,
but extended the models to accommodate unequal variances for
random effects where required.
In studies reporting multiple tests in the same participants, results R E S U L T S
were expressed separately for each test component.

Results of the search
Investigations of heterogeneity
The results of electronic database and handsearching are outlined
Factors that could influence diagnostic accuracy other than true in Figure 1. There were no disagreements between authors about
test performance included those relating to methodological quality either the number of studies eligible for inclusion, nor data re-
and study design, characteristics of the underlying population, and sults (κ = 1.0). We identified 26 reports of 25 studies (35 compar-
characteristics of the index and reference test. We detailed and isons in total). Seven studies compared more than one test versus
compared patient inclusion criteria for each included study. We coronary angiography, and were interrogated to contribute data
also investigated heterogeneity statistically by: to more than one test comparison (De Lima 2003; Gang 2007;
1. applying separate models to different subgroups Garcia-Canton 1998; Garg 2000; Sharma 2005; Sharma 2009;
2. adding covariates to the hierarchical model. Vandenberg 1996). One study was reported twice (Sharma 2005),
Factors such as differences in study population characteristics (e.g. and one study (Sharples 2004) could not contribute to the meta-
prevalence of chest pain, hypertension and diabetes) and test ap- analysis because it reported results per coronary vessel, but not per
plication (diagnostic test threshold, criteria for positive test, choice patient. The diagnostic and treatment pathway is presented at the
of stress agent and stress protocol, and operator variability) were patient level, but including vessel-level analysis lead to inappro-
used to explore any heterogeneity discovered in the analysis for priate weighting in the combined analysis, and the potential for
each test separately, and to assess the impact of heterogeneity on bias from clustering of patients’ results. The details of all studies
the relative accuracy across tests. included in the meta-analysis are reported in Characteristics of
For index tests such as ECG and echocardiography, where different included studies and Table 2.

Cardiac testing for coronary artery disease in potential kidney transplant recipients (Review) 6
Copyright © 2011 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd.