 |
|
||||
| Home | Login | |||||
| About | Editorial board | Articles
|
NSI Publications
|
Search | Instructions | Online Submission | Subscribe | Videos | Etcetera | Contact |
|
||||||||||||||||||||
|
|
|
Accuracy of computed tomography perfusion in detecting delayed cerebral ischemia following aneurysmal subarachnoid hemorrhage: A meta-analysis
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.121922
Background and Purpose: In recent years, significant literature shows that computed tomography perfusion (CTP) can provide sufficient information on cerebral hemodynamics and effectively indicate delayed cerebral ischemia (DCI) before the development of infarction. We aimed at performing a meta-analysis to provide a more full and accurate evaluation of CTP and CTP parameters in detecting DCI in patients with aneurysmal subarachnoid hemorrhage. Materials and Methods: We searched the PubMed, MedLine, Embase and Cochrane databases for analysis published from February 2005 to February 2013. We extracted CTP parameters, including cerebral blood volume (CBV), cerebral blood flow (CBF), mean transit time (MTT), time to peak (TTP), interhemispheric ratios for CBV and CBF and interhemispheric differences for MTT and TTP. Pooled estimates of sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR) and the summary receiver-operating characteristic curve were determined. Results: Four research studies are met the inclusion criteria for the analysis. The pooled sensitivity, specificity, PLR, NLR and DOR of CTP for detecting the DCI were 82%, 82%, 4.56, 0.22 and 20.96, respectively. Through the evaluation of absolute CTP parameters, CBF and MTT showed diagnostic value for DCI, but CBF and TTP did not. Moreover, CBF ratio, MTT difference and TTP difference showed more diagnostic value than CBV ratio in DCI detection by the assessment of relative CTP parameters. Conclusions: As a non-invasive and short time consuming screening method, CTP own a high diagnostic value for the detection of DCI after aneurysm rupture. Keywords: Aneurysmal subarachnoid hemorrhage, computed tomography perfusion, delayed cerebral ischemia
Delayed cerebral ischemia (DCI) is a devastating complication and leads to significant morbidity and mortality in this patients population following aneurysmal subarachnoid hemorrhage (aSAH), DCI typically develops between 4 and 14 days after aneurysm rupture. [1],[2] In clinical practice, unexplained clinical deterioration and/or new infarction on image examinations may be diagnosed as DCI after aneurysm rupture. However it is limited to effectively improve clinical outcomes by treatment interventions once infarction has occurred. [3] In addition, as a controversial treatment, hyperdynamic therapy is instituted to treat DCI and the serious complication may increase the risk of morbidity and mortality of aSAH patients if the evaluation of DCI is uncertain. [4],[5],[6] Therefore, diagnostic techniques, which can provide sufficient information on cerebral hemodynamics and effectively indicate DCI before the development of infarction, are very needed. Recently, a non-invasive and less consuming screening method, computed tomography perfusion (CTP) is used to assess the changes of cerebral blood flow (CBF) and indicate DCI. [7],[8] Moreover, studies [1],[4] had suggested that CTP can predict DCI with high sensitivity and specificity. The meta-analysis meant to provide a more full and accurate evaluation of CTP and CTP parameters in detecting DCI after aneurysm rupture.
Literature search strategy The PubMed, MedLine, Embase and Cochrane were searched from February 2005 to February 2013 by two authors (H.G.S. and H.M.Z.). In the search, all articles that evaluated the diagnostic performance of CT perfusion for DCI in patients with aSAH were identified and the terminology about designing the electronic search included "CTP" or "CT perfusion" or "computed tomography perfusion", "DCI" or "delayed cerebral ischemia" or "ischemia", and "subarachnoid haemorrhage" or "subarachnoid hemorrhage". All published articles printed in English and Chinese language and the relevant studies were also searched from the reference lists of all relevant articles. Study selection and inclusion criteria Two authors (H.G.S. and H.M.Z.) independently and closely evaluated the eligible articles on the basis of the titles, abstracts and keywords and excluded clearly irrelevant studies. Then, three authors (H.G.S., H.M.Z., J.P.M.) independently assessed the full text of the remaining studies and selected the articles that met the inclusion criteria of this meta-analysis. The inclusion criteria were: (1) SAH was caused by ruptured aneurysm; (2) the aneurysms had been repaired by surgical clipping and/or endovascular coiling; (3) all patients underwent CTP before the treatment of DCI; (4) the study could provide adequate information to rebuild fourfold table; (5) all patients could not be reported in other publications. Studies were excluded when they have not met the above-mentioned inclusion criteria. Any disagreements were resolved by consensus. Data extraction DCI is defined as clinical deterioration as manifested by new neurologic deficits or worsening on Glasgow Coma Scale. As possible causes of clinical deterioration, aneurysm rebleeding, intracranial hemorrhage, seizure activity, hydrocephalus, infection and metabolic disturbance are not included in this definition. Some of the authors [3],[9],[10] believe that new infarction on image examinations was another component of DCI definition and it was also included. Absolute CTP parameters included cerebral blood volume (CBV), CBF, mean transit time (MTT) and time to peak (TTP). Some studies calculated the optimal threshold values and corresponding sensitivity and specificity, because there were no standardized thresholds for CTP to determine DCI. [1],[9],[10],[11] Moreover, as relative CTP parameters, interhemispheric ratios for CBV and CBF and interhemispheric differences for MTT and TTP were also calculated in some studies. [10],[11] All the above-mentioned CTP parameters were extracted in this meta-analysis [Table 1].
As the scoring criteria, the quality assessment of diagnostic accuracy studies (QUADAS) tool [12],[13] was used to evaluate the methodologic quality of each enrolled study, such as study design, bias and limitations. The QUADAS tool includes 14 items and each item can be scored as "yes", "no", or "unclear." We determined that all items were weighted equally and scored 1 for "yes", 0 for "no" and 0.5 for "unclear." As the statistical data, true positive, false positive, true negative and false negative were extracted. However, if these data were not reported, we can calculate it by using other mentioned data, such as sensitivity and specificity. Extracted information also included information of patients, study design, type of CT scanner and QUADAS score. Two authors (H.M.Z., J.P.M.) performed the above data extraction, calculation and evaluation on each research study and disagreements were resolved by consensus after discussion [Table 1], [Table 2] and [Table 3].
Statistical analysis The pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR) and diagnostic odds ratio (DOR) with the 95% of confidence intervals (CIs) were calculated. Cochran-Q tests, I2 and Chi-square tests were performed to evaluate the heterogeneity between studies. If the heterogeneity was not statistically significant, the pooled sensitivities and specificities and DOR were calculated by the fix-effect model of Mantel and Haenszel method. In addition, the random-effect model of DerSimonian and Laird method was performed when heterogeneity was statistically significant. Forest plots were painted to display the variations of sensitivities and specificities from different studies. In the right conditions, summary receiver-operating characteristic (SROC) curve would be a draw and the area under the SROC curve with its Q* point would be calculated too. The above mentioned statistical analyses were performed by using Meta-Disc 1.4 (Clinical Biostatistics Unit, Hospital Ramon y Cajal, Madrid, Spain).
From the initial 96 articles, 88 articles were considered as irrelevant for this analysis and exclude by assessed their titles, abstracts and keywords. Full text of the 8 relevant articles were obtained and evaluated. Only 4 of these articles contained the appropriated data and were included the analysis. A total of 2 articles could not get enough data to construct fourfold table and were not included. [14],[15] The remaining 2 articles were excluded because the data from the same sample [Figure 1]. [3],[9] Due to the small sample size in this analysis, the SROC curve was not draw and the area under the SROC curve with its Q* point was no possible to calculate.
Quantitative analysis of computed tomography perfusion A total of 135 patients included in the CTP analysis and 73 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of CTP for diagnosing DCI after aSAH were 82% (95% CI: 0.71-0.90), 82% (95% CI: 0.70-0.91), 4.56 (95% CI: 2.65-7.88), 0.22 (95% CI: 0.13-0.36) and 20.96 (95% CI: 8.88-49.48), respectively [Table 4]. The value of Cochran's Q of DOR was 0.01 (P = 0.9063) and I2 was 0.0%, indicating that there was not statistically significant between-study heterogeneity of DOR in these data [Figure 2].
Quantitative analysis of computed tomography perfusion parameters Cerebral blood flow A total of 181 patients included in the CBF analysis and 98 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal CBF threshold values for detecting DCI after aSAH were 76% (95% CI: 0.66-0.84), 67% (95% CI: 0.56-0.77), 2.28 (95% CI: 1.65-3.15), 0.36 (95% CI: 0.24-0.53) and 6.34 (95% CI: 3.38-11.91), respectively [Table 4]. The value of Cochran's Q of DOR was 0.65 (P = 0.4207) and I2 was 0.0%, indicating that there was not statistically significant between-study heterogeneity of DOR in these data. Mean transit time Ninety-one patients were diagnosed with DCI in the MTT analysis with 181 patients. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal MTT threshold values for detecting DCI after aSAH were 71% (95% CI: 0.61-0.80), 73% (95% CI: 0.62-0.82), 2.65 (95% CI: 1.81-3.87), 0.40 (95% CI: 0.28-0.55) and 6.71 (95% CI: 3.57-12.62), respectively [Table 4]. The value of Cochran's Q of DOR was 0.18 (P = 0.6708) and I2 was 0.0%, indicating that there was not statistically significant between-study heterogeneity of DOR in these data. Cerebral blood volume ratio 131 patients included in the CBV ratio analysis and 66 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal CBV ratio threshold values for detecting DCI after aSAH were 67% (95% CI: 0.54-0.78), 66% (95% CI: 0.53-0.77), 1.95 (95% CI: 1.34-2.84), 0.49 (95% CI: 0.33-0.75) and 3.89 (95% CI: 1.94-7.80), respectively [Table 4]. The value of Cochran's Q of DOR was 1.02 (P = 0.3132) and I2 was 1.7%, indicating that there was not statistically significant between-study heterogeneity of DOR in these data. Cerebral blood flow ratio The CBF ratio analysis involved 131 patients and 66 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal CBF ratio threshold values for detecting DCI after aSAH were 76% (95% CI: 0.64-0.85), 77% (95% CI: 0.65-0.86), 4.27 (95% CI: 0.76-23.84), 0.34 (95% CI: 0.22-0.55) and 12.85 (95% CI: 1.75-94.63), respectively [Table 4]. The value of Cochran's Q of DOR was 3.86 (P = 0.0496) and I2 was 74.1%, indicating the between-study heterogeneity of DOR was statistically significant. Mean transit time difference The analysis of MTT difference involved 131 patients and 66 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal MTT difference threshold values for detecting DCI after aSAH were 79% (95% CI: 0.67-0.88), 66% (95% CI: 0.53-0.77), 2.25 (95% CI: 1.56-3.25), 0.33 (95% CI: 0.20-0.55) and 6.85 (95% CI: 3.21-14.61), respectively [Table 4]. The value of Cochran's Q of DOR was 0.00 (P = 0.9689) and I2 was 0.0% and it demonstrated the between- study heterogeneity of DOR was not statistically significant. Time to peak difference The TTP difference analysis involved 131 patients and 66 patients were diagnosed with DCI. The pooled sensitivity, specificity, PLR, NLR and DOR of optimal TTP difference threshold values for detecting DCI after aSAH were 71% (95% CI: 0.59-0.82), 77% (95% CI: 0.65-0.86), 2.74 (95% CI: 1.77-4.26), 0.38 (95% CI: 0.25-0.59) and 6.94 (95% CI: 3.33-14.47), respectively [Table 4]. The value of Cochran's Q of DOR was 3.49 (P = 0.0617) and I2 was 71.4% and it demonstrated that there was moderate between-study heterogeneity of DOR in these data. cerebral blood volume and time to peak There were no statistically significant differences in the mean values of CBV and TTP between patients with and without DCI after aSAH and the sensitivity and specificity of optimal CTP threshold values were not calculated in some studies. [1],[10],[11] Only one article showed that optimal threshold values were 2.78 mL/100 g (52% sensitivity, 63% specificity) for CBV and 25.2 s (54% sensitivity, 63% specificity) for TTP [Table 4]. [11] Therefore, the absolute parameters of CBV and TTP were not enrolled in this meta-analysis.
In this analysis, CTP, which had 82% (71-90%) sensitivity, 82% (71-91%) specificity and 20.96 (8.88-49.48) DOR, is confirmed to own a high diagnostic value for the detection of DCI in patients with aSAH. Through the evaluation of absolute CTP parameters, CBF and MTT showed diagnostic value for DCI but CBF and TTP did not. Moreover, CBF ratio, MTT difference and TTP difference showed more diagnostic value than CBV ratio in DCI detection by the assessment of relative CTP parameters. There were some causes might explain the false-positive and false-negative findings in including studies. First, other pathological conditions might blur the results of perfusion measurements and it contributed to the production of the false-positive findings. The presence of hydrocephalus, edema and infection might coexist with DCI and strengthen the influence of cerebral perfusion in the scanned regions, although these conditions caused clinical deterioration could not be a standard in the definition of DCI. [4],[11] Second, DCI could be transient and it might miss a positive result when the CTP was performed after the improvement of neurological condition. [11] Third, some institutions might not use modern multi-detector CT for these studies and some small size abnormalities were hardly detected in the scanned regions. [4] Fourth, perfusion measurements indicated that there were some patients with bilateral ischemia and false-negative CTP findings could occur when used the relative CTP parameters to assess DCI. [11] False negative CTP findings can also result from the pre-defined scanned regions. To minimize observer bias, some institutions [9],[11] pre-defined the scanned regions for the measurements of perfusion. The pre-defined scanned regions might not always include the area of ischemia and differences of measurements might be less conspicuous. Konstas et al. [16] believed the relative CTP parameters had better diagnostic properties than the absolute CTP parameters, because the absolute CTP parameters were readily influenced by venous output scaling factor and existed potential variability. Dankbaar et al. [11],[17] also favored the relative CTP parameters rather the absolute one and they insisted that the use of relative CTP parameters reduced the variability caused by post-processing steps. However, there were some patients with bilateral and diffuse ischemia after aSAH and the interhemispheric differences of measurements might be not significant. Therefore, the absolute CTP parameters owned better diagnostic properties than the relative CTP parameters in this patient population. [9],[11] Some limitations should be considered in our meta-analysis. First, we must emphasize the fact that we have tried to extensive search the relevant research studies for this analysis, but it hard to find a new one. Due to the small sample size in this analysis, the SROC curve was not draw and the area under the SROC curve with its Q* point was no possible to calculate. When more relevant studies are published in the future, we will perfect the analysis without delay. Second, in Van' study, DCI is defined as clinical deterioration and/or new infarction on image examinations while DCI is defined as clinical deterioration only in another three studies. Variability occurred in the analysis, because of the difference in DCI definition. Third, due to there were no standardized thresholds for CTP to determine DCI, the difference in optimal threshold values of each CTP parameters led to the existence of bias.
This article is supported by the national natural science foundation of china (No. 30801185)
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4]
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||
