Optimal Timing of Diffusion-Weighted Imaging to Avoid False-Negative Findings in Patients With Transient Ischemic Attack
This article has a correction. Please see:
Background and Purpose—We aimed to investigate the optimal timing of diffusion-weighted imaging (DWI) in patients with transient ischemic attack (TIA).
Methods—Seventy-three consecutive patients with TIA underwent DWI on admission (initial DWI) and at 24 hours after admission (second DWI). Patients were divided into 2 groups based on initial DWI findings in relation to the second examination: false negative (group 1) and other (group 2). The probability of initial false-negative findings was determined for each hour from TIA onset to initial DWI. Multivariate analysis was used to evaluate the independent risk factors associated with false-negative findings on initial DWI.
Results—Of the 73 patients examined (56 men; mean age, 68 years), 9 (12%) were categorized into group 1. The latency from TIA onset to initial DWI was 1.7±0.6 hours for group 1 (range, 1–2.8 hours) and 3.3±2.6 hours for group 2 (range, 35 minutes to 12 hours). The probability of false-negative findings on initial DWI decreased in a time-dependent manner (25%, 21%, and 7% for 1, 2, and 3 hours, respectively), and no false-negative findings were observed on initial DWI performed at >3 hours from symptom onset. Short latency (≤2 hours) from TIA onset to initial DWI was an independent risk factor related to false-negative findings (odds ratio, 13.11; 95% confidence interval, 1.07–161.38; P=0.045).
Conclusions—If the duration between TIA symptom onset and initial DWI is <2 hours, a repeat examination should be performed to minimize the risk of false-negative findings.
Diffusion-weighted imaging (DWI) is useful for detecting acute cerebral ischemic lesions and diagnosing transient ischemic attack (TIA); however, DWI may not necessarily provide positive findings, depending on the time-based TIA definition. TIA with DWI-positive findings is TIA with infarction,1 and patients with evidence of DWI-positive lesions are at a higher risk of recurrent stroke.1–4 Thus, whether a TIA involves infarction or not is one of the most important factors determining a patient’s prognosis. Unfortunately, false-negative findings on initial DWI are common, especially in patients with a short latency from symptom onset to initial DWI.5–8
For instance, Morita et al9 reported that, among 31 patients with TIA who initially had negative DWI findings, a hyperintense lesion was found on a second scan in 10 individuals (32.3%), and the sensitivity of DWI for patients with TIA was better within 24 hours of onset than within 3 hours. Although several risk factors related to false-negative findings on initial DWI have been reported,5,10,11 to our knowledge, no detailed analysis has focused on the relationship between the latency from symptom onset to initial DWI (DWI latency) obtained using a 3-T magnetic resonance imaging (MRI) device and false-negative findings on initial DWI.
Determining the optimal timing of DWI in patients with TIA may be a key to making an accurate prognosis. We hypothesized that a short DWI latency would be an independent risk factor for false-negative findings. Our aim was to elucidate the optimal timing of DWI to minimize false-negative findings in patients with TIA.
Patients and Methods
This study was approved by the institutional ethics committee. A brief overview of the study methodology is given below, whereas the full description is available in the online-only Data Supplement.
Seventy-three consecutive patients, who were admitted to the stroke care unit at our hospital between January 2011 and August 2015 and were clinically diagnosed with definite TIA, were enrolled. All patients immediately underwent an MRI examination on arrival. The diagnosis of TIA was made by a team of neurosurgeons, neurologists, and neuroradiologists, and all patients underwent neurological evaluation by a stroke specialist. The initial MRI scan, which obtained diffusion-weighted images (slice thickness, 5 mm; gap, 1 mm; Discovery MR 750; GE Healthcare, Milwaukee, WI), was followed by a second examination 24 hours later, using a previously reported method.9
Patients were divided into 2 groups based on findings of the initial and second DWI examinations (Figure I in the online-only Data Supplement). In group 1, lesions were not found on initial DWI but were found on the second scan. In group 2, both examinations were negative or both were positive. The false-negative rate was determined for each 1-hour interval after symptom onset. The relationship between false-negative findings and DWI latency was determined using multivariate analysis.
All continuous variables are expressed as the mean±SD. The χ2 test, Student t test, and Mann–Whitney U test were used to compare nominal, parametric numeric, and nonparametric numeric variables, respectively. Cutoff points for nonparametric numeric variables were determined using the receiver operating characteristic curve. For all statistical analyses, P<0.05 was considered to represent a significant difference.
Of 73 patients (56 men; age, 68.4±13.7 years), 25 (34%) had positive findings on initial DWI, 39 (54%) did not have positive findings on either scan, and 9 patients (12%) did not have positive findings on initial DWI but had such findings on the second scan. Thus, group 1 included 9 patients, and group 2 included 64 patients (Figure I in the online-only Data Supplement).
Table I in the online-only Data Supplement shows the patients’ characteristics, risk factors, medical history, initial symptoms, and TIA cause. Initial DWI was performed at 1 to 2.8 hours (1.7±0.6 hours) after onset in group 1, and at 35 minutes to 12 hours (3.3±2.6 hours) in group 2 (Figure II in the online-only Data Supplement).
The probability of false-negative findings for each 1-hour interval was 25% (2/8), 21% (6/28), 7% (1/15), and 12% (9/73) for 0 to 1, 1 to 2, 2 to 3, and 0 to 12 hours from symptom onset, respectively (Figure III in the online-only Data Supplement). This probability tended to decrease in a time-dependent manner, and no false-negative findings were observed on DWI performed after 3 hours.
Using receiver operating characteristic curve analysis, the cutoff points for DWI latency and National Institutes of Health Stroke Scale score on admission were 2 hours and 2 points, respectively (Figure IV in the online-only Data Supplement). Only variables with a P value <0.1 in univariate analysis were included in the multivariate analysis. The DWI latency (odds ratio, 13.11; 95% confidence interval, 1.07–161.38; P=0.045), National Institutes of Health Stroke Scale score on admission (odds ratio, 9.95; confidence interval, 1.03–95.81; P=0.047), and history of TIA (odds ratio, 18.07; confidence interval, 1.62–201.24; P=0.019) were independent risk factors for false-negative findings (Table).
In this retrospective study, we demonstrated that a short latency (within 2 hours) of DWI performed using a 3-T MRI device was an independent risk factor for false-negative findings.
Chalela et al10 reported that the false-negative rate of 1.5-T MRI findings on admission within 3 hours from onset was 27% and decreased in a time-dependent manner. Oppenheim et al5 reported no false-negative initial DWI in patients examined after 24 hours, whereas an initial DWI scan performed using a 1.5-T MRI device was associated with false-negative findings in patients examined within 24 hours TIA or acute ischemic stroke onset. Although patients in our study were examined via 3-T MRI with shorter DWI latency than that noted in the aforementioned studies, our results are consistent with the findings of these reports. However, the discrepancy in the size of the 2 groups may have biased our results.
We considered that TIA with false-negative findings on initial DWI corresponds to TIA with delayed infarction, which should be categorized as TIA with infarction. Although previous reports have shown that DWI-positive TIA is associated with a higher risk of recurrent stroke than that of non-DWI–positive lesions,1–4 some of our patients with false-negative findings who might have been included in a DWI-positive TIA group had the DWI latency been longer, as in the previous studies.
To our knowledge, this is the first study to determine the DWI latency associated with false-negative findings when using 3-T MRI; performing an examination too early may mean that some DWI-positive patients are not identified. Indeed, negative TIA findings on early DWI should be interpreted with caution, and, in the absence of an alternative diagnosis, patients with clinical TIA but negative DWI findings should be assumed to have TIA. We would like to emphasize that we do not mean to suggest that initial DWI should be delayed until >3 hours after symptom onset; instead, we recommend that DWI should be repeated in patients with negative findings on initial DWI. When examining patients with suspected TIA, physicians should consider the possibility of false-negative findings on DWI performed within 2 hours from symptom onset. A false-negative finding is only important if the physician suspects a different diagnosis, such as migraine, where stroke prevention medications need not be administered.
The possibility of false-negative findings tends to decrease every hour within the first 3 hours from onset; no false-negative findings are expected after 3 hours. A short DWI latency (≤2 hours) is an independent risk factor for an initial false-negative DWI. Repeat DWI is recommended to avoid underestimating the risk of recurrent TIA or stroke.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.014576/-/DC1.
- Received January 11, 2017.
- Revision received March 14, 2017.
- Accepted April 5, 2017.
- © 2017 American Heart Association, Inc.
- Arsava EM,
- Furie KL,
- Schwamm LH,
- Sorensen AG,
- Ay H
- Purroy F,
- Montaner J,
- Rovira A,
- Delgado P,
- Quintana M,
- Alvarez-Sabín J
- Redgrave JN,
- Coutts SB,
- Schulz UG,
- Briley D,
- Rothwell PM
- Oppenheim C,
- Stanescu R,
- Dormont D,
- Crozier S,
- Marro B,
- Samson Y,
- et al
- Lövblad KO,
- Laubach HJ,
- Baird AE,
- Curtin F,
- Schlaug G,
- Edelman RR,
- et al
- Boulanger JM,
- Coutts SB,
- Eliasziw M,
- Subramaniam S,
- Scott J,
- Demchuk AM
- Zuo L,
- Zhang Y,
- Xu X,
- Li Y,
- Bao H,
- Hao J,
- et al