Comatose Patients with Cardiac Arrest Predicting Clinical Outcome

Document Sample
Comatose Patients with Cardiac Arrest Predicting Clinical Outcome Powered By Docstoc
					                 Note: This copy is for your personal non-commercial use only. To order presentation-ready
                 copies for distribution to your colleagues or clients, contact us at

                                                                                                                                                        ORIGINAL RESEARCH
                                                             Comatose Patients with Cardiac
                                                             Arrest: Predicting Clinical Outcome
                                                             with Diffusion-weighted MR Imaging1

Ona Wu, PhD
                                                                         Purpose:    To examine whether the severity and spatial distribution
A. Gregory Sorensen, MD
                                                                                     of reductions in apparent diffusion coefficient (ADC) are
Thomas Benner, PhD
                                                                                     associated with clinical outcomes in patients who become
Aneesh B. Singhal, MD                                                                comatose after cardiac arrest.
Karen L. Furie, MD
David M. Greer, MD                                                   Materials and   This was an institutional review board–approved, HIPAA-
                                                                         Methods:    compliant retrospective study of 80 comatose patients
                                                                                     with cardiac arrest who underwent diffusion-weighted
                                                                                     magnetic resonance imaging. The need to obtain informed
                                                                                     consent was waived except when follow-up phone calls
                                                                                     were required; in those cases, informed consent was ob-
                                                                                     tained from the families. Mean patient age was 57 years
                                                                                     16 (standard deviation); 31 (39%) patients were women.
                                                                                     ADC maps were semiautomatically segmented into the
                                                                                     following regions: subcortical white matter; cerebellum;
                                                                                     insula; frontal, occipital, parietal, and temporal lobes; cau-
                                                                                     date nucleus; putamen; and thalamus. Median ADCs were
                                                                                     measured in these regions and in the whole brain and were
                                                                                     compared (with a two-tailed Wilcoxon test) as a function
                                                                                     of clinical outcome. Outcome was defined by both early
                                                                                     eye opening in the 1st week after arrest (either spontane-
                                                                                     ously or in response to external stimuli) and 6-month
                                                                                     modified Rankin scale score.

                                                                          Results:   Whole-brain median ADC was a significant predictor of
                                                                                     poor outcome as measured by no eye opening (specificity,
                                                                                     100% [95% confidence interval {CI}: 86%, 100%]; sensi-
                                                                                     tivity, 30% [95% CI: 18%, 45%]) or 6-month modified
                                                                                     Rankin scale score greater than 3 (specificity, 100% [95%
                                                                                     CI: 73%, 100%]; sensitivity, 41% [95% CI: 29%, 54%]),
                                                                                     with patients with poor outcomes having significantly
                                                                                     lower ADCs for both outcome measures (P         .001). Dif-
                                                                                     ferences in ADC between patients with good and those
                                                                                     with poor outcomes varied according to brain region, in-
                                                                                     volving predominantly the occipital and parietal lobes and
                                                                                     the putamen, and were dependent on the timing of imag-
 From the Athinoula A. Martinos Center for Biomedical
Imaging, Department of Radiology, Massachusetts Gen-                   Conclusion:   Spatial and temporal differences in ADCs may provide
eral Hospital, 149 13th St, CNY 2301, Charlestown, MA                                insight into mechanisms of hypoxic-ischemic brain injury
02129 (O.W., A.G.S., T.B.); and Department of Neurology,
                                                                                     and, hence, recovery.
Massachusetts General Hospital, Boston, Mass (A.B.S.,
K.L.F., D.M.G.). Received July 21, 2008; revision re-
quested September 18; revision received December 9;                                   RSNA, 2009
accepted December 23; final version accepted January
28, 2009. Supported in part by the Medical Investigation
of Neurodevelopmental Disorders, or MIND, Institute.
Address correspondence to O.W. (e-mail: ona@nmr ).

    RSNA, 2009

Radiology: Volume 252: Number 1—July 2009 ▪                                                                                173
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                                  Wu et al

   n comatose patients after cardiac             nosis for the majority of patients                 The purpose of our study was to
   arrest, prediction of meaningful              whose conditions do not fit into either        examine whether the severity and spa-
   neurologic recovery within the first           the very good or very grim categories.        tial distribution of reductions in ADC
several weeks can be challenging.                    Neuroimaging is potentially useful        are associated with clinical outcomes in
Mortality rates with cardiac arrest              in the prediction of neurologic recov-        patients who become comatose after
range from 60% to 85%, and approx-               ery; however, literature to support the       cardiac arrest.
imately 80% of survivors are initially           use of neuroimaging is quite sparse
comatose (1–3). Of those who survive,            and is limited by small sample sizes
50% are left with a permanent neuro-             and infrequent quantification of the             Materials and Methods
logic disability, and only 10% are able          findings. Computed tomographic (CT)
to resume their former lifestyle (1,2).          findings are most often normal when            Patients
Multiple methods have been assessed              CT is performed early after cardiac           We retrospectively analyzed data ac-
for prognostic value in this setting, in-        arrest (10), and patients who develop         quired as part of a Massachusetts
cluding the clinical examination (3),            dramatic changes at CT typically do           General Hospital Institutional Review
electrophysiology with electroen-                not pose diagnostic dilemmas, thus            Board–approved, Health Insurance
cephalography (4) and somatosensory              leaving CT as a technique of limited          Portability and Accountability Act–
evoked potentials (5), and plasma bi-            value in most cases. Magnetic reso-           compliant, single-center prospective
omarkers indicative of neuronal cell             nance (MR) imaging remains a more             observational study of 500 patients
death (6). Some groups (7) have                  promising technique because of its            with nontraumatic coma. Data were
looked at a combined approach with               sensitivity at earlier time stages. Dif-      collected from 2000 to 2007. Within
these modalities to improve the power            fusion-weighted (DW) imaging ap-              this patient cohort, 200 patients were
of prognostication. However, prog-               pears particularly promising, as it is        comatose after cardiac arrest. The de-
nostication can be achieved with very            highly sensitive to early ischemic            cision to perform additional testing,
high specificity only for patients who            changes (11). Experimental animal             including neuroimaging, was at the
rapidly awaken after their arrest and            models of global ischemia showed that         discretion of the treating clinicians. A
are destined for a good outcome (8)              changes in the apparent diffusion co-         total of 80 patients (age range, 23– 88
and for those who remain deeply co-              efficient (ADC) were immediate (12)            years; mean age, 57 years 16 [stan-
matose at 72 hours with minimal                  and varied over time and by tissue            dard deviation]) consisting of 49 men
brainstem activity or absent cortical            region (13). Furthermore, DW imag-            (age range, 23–79 years; mean age, 57
somatosensory evoked potentials and              ing abnormalities in patients with car-       years 16) and 31 women (age range,
are destined for a poor outcome (9).             diac arrest have correlated with poor         23– 88 years; mean age, 57 years
Physicians are often left in a quandary          outcome in several small studies ( 12         16) underwent MR imaging performed
trying to predict the neurologic prog-           patients) (14), and it appears that
                                                 changes may occur in different brain
                                                 regions on the basis of the timing of          Published online before print
 Advances in Knowledge                           imaging after arrest (15) or the nature        10.1148/radiol.2521081232
      Whole-brain median ADC was a               of the injury (eg, primary hypoxia fol-        Radiology 2009; 252:173–181
      significant predictor of poor out-          lowed by ischemia) (16).
      come as measured by no eye                                                                Abbreviations:
                                                                                                ADC apparent diffusion coefficient
      opening in the 1st week after car-
      diac arrest or a modified Rankin             Implications for Patient Care                 DW diffusion weighted
                                                                                                IQR interquartile range
      score greater than 3, and patients             Serial MR imaging may allow cli-
      with poor outcomes had signifi-                 nicians to stage the degree of re-         Author contributions:
                                                                                                Guarantors of integrity of entire study, O.W., D.M.G.;
      cantly lower ADCs for both met-                covery or injury in comatose pa-
                                                                                                study concepts/study design or data acquisition or data
      rics.                                          tients with cardiac arrest, poten-         analysis/interpretation, all authors; manuscript drafting or
      ADC differences between patients               tially improving the ability of            manuscript revision for important intellectual content, all
      with good and those with poor                  clinicians to predict which pa-            authors; manuscript final version approval, all authors;
      outcomes varied according to                   tients have a chance for a good            literature research, O.W., A.B.S., D.M.G.; clinical studies,
      brain region and timing of imag-               outcome.                                   O.W., A.G.S., T.B., A.B.S., K.L.F., D.M.G.; statistical analy-
      ing.                                           Evaluating patients for severe             sis, O.W., K.L.F., D.M.G.; and manuscript editing, all
      Multivariate analysis of ADCs in               whole-brain ADC reductions can
      occipital and parietal lobes and               be used for confirmatory testing,           Funding:
      putamen increased sensitivity in               thereby potentially avoiding long          This research was supported in part by the National Insti-
      predicting poor outcome while                  and futile support of patients with        tutes of Health (grants R01NS59775, R01NS38477,
                                                                                                P50NS051343, R01NS051412, and NCRR [P41RR14075]).
      maintaining high specificity com-               a high likelihood of a poor
      pared with univariate models.                  outcome.                                   Authors stated no financial relationship to disclose.

174                                                                          ▪ Radiology: Volume 252: Number 1—July 2009
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                             Wu et al

by using a 1.5-T imaging unit. Eight                         brain damage by using a quantitative           potential sign of poor recovery in patients
patients underwent two MR imaging                            and automated approach. ADC maps               who undergo life support withdrawal
examinations, and one underwent                              were calculated from the slope of the          (and who thus would die without the po-
three. For these eight patients, we                          linear regression fit of the log of the         tential for demonstrating recovery at 6
used only the initial studies for analy-                     DW images (b value 1000 sec/mm2)               months). We performed a similar analy-
sis. Fourteen patients underwent in-                         and the images with a b value of 0             sis (with two-tailed Wilcoxon rank-sum
duced hypothermia treatment within 6                         sec/mm2. The latter images were non-           tests) by using a modified Rankin scale
hours of cardiac arrest. The require-                        linearly coregistered (19) to the              score greater than 3 at 6 months to rep-
ment to obtain informed consent was                          ICBM-452 T1 5th Order Polynomial               resent poor outcome and a modified
waived unless the patient survived the                       Warps Atlas (20). By using the Inter-          Rankin scale score of 3 or less to repre-
hospital admission and a follow-up                           national Consortium for Brain Map-             sent good outcome.
phone call was required at 3 and/or 6                        ping probabilistic atlases (21,22),                The Fisher exact two-tailed test was
months; in these cases, informed con-                        masks were generated by using a                used to test for significant differences be-
sent was obtained from the families.                         threshold of 50% (Fig 1) for the fol-          tween patients who were treated with in-
                                                             lowing regions: subcortical white mat-         duced hypothermia and those who were
Imaging Studies                                              ter; cerebellum; insula; frontal, occip-       not (JMP, version 7.0). Subset analysis to
Patients underwent MR imaging per-                           ital, parietal, and temporal lobes; cau-       examine differences between early and
formed by using 1.5-T imaging units                          date nucleus; putamen; and thalamus.           late imaging was performed for patients
(GE Medical Systems, Milwaukee,                              Median ADC was automatically mea-              imaged at less than 3 days and those im-
Wis). Owing to the retrospective na-                         sured in these regions, as well as in          aged at 3 or more days after cardiac ar-
ture of this study, the DW imaging                           the entire brain, by using these masks.        rest because the median time to the first
protocol changed over time. Repeti-                          To minimize effects from cerebral spi-         MR imaging examination was 2 days
tion time was typically 7500 msec (n                         nal fluid, analysis was limited to ADCs         (JMP, version 7.0). Backward stepwise
48) or 5000 msec (n        30), with one                     of 1200     10 6 mm2/sec or less (23).         logistic regression was performed to in-
patient imaged by using a repetition                                                                        vestigate the relationship between Glas-
time of 10 000 msec and another im-                          Statistical Analysis                           gow Coma Scale score at admission (ob-
aged by using a repetition time of 6300                      Spatial differences among the regions          tained on day 0 or day 1), age, sex, time
msec. A minimum echo time, typically                         were examined by using a one-way anal-         to MR imaging, hypothermia therapy ad-
between 71 and 103 msec, was used.                           ysis of variance with a post-hoc Student-      ministration, and ADC with 6-month
Axial images were acquired with sec-                         Newman-Keuls test (KaleidaGraph, ver-          modified Rankin scale scores dichoto-
tion thicknesses ranging from 5 to 6                         sion 4.0; Synergy Institute, Reading, Pa).     mized as indicating poor (modified
mm and a 1-mm intersection gap. A                            Median ADC in the regions, as well as in       Rankin scale score      3) or good (modi-
128 128 acquisition matrix was used                          the whole brain, was compared with clin-       fied Rankin scale score         3) outcome
with 37 studies zero filled to 256                            ical outcome defined by the modified             (JMP, version 7.0). Sensitivity values
256, resulting in in-plane resolutions                       Rankin scale score and eye opening. Pa-        were calculated for the case of no false-
ranging from 0.86          0.86 mm 2 to                      tients underwent routine clinical evalua-      positive findings and the maximum num-
1.9 1.9 mm . Multisection axial DW                           tions, including neurologic assessments,       ber of true-positive findings. Ninety-five
images were acquired by sampling the                         on days 0, 1, 3, and 7 after cardiac arrest,   percent confidence intervals were calcu-
full diffusion tensor by using either                        and a modified Rankin scale score was           lated according to the efficient-score
single-shot pulsed field gradient spin-                       obtained at 6 months in surviving pa-          method (corrected for continuity)
echo echo-planar imaging sequences                           tients. Because this was an observational      (24,25). The sensitivity of the univariate
with post hoc eddy current correction                        study and the investigators were not in-       analysis was compared with that of the
(17) or a twice-refocused spin-echo echo-                    volved in the decision to continue or with-    multivariate analysis by using a 2 test
planar imaging sequence (18). The total                      draw care, many patients underwent             (26) (R, version 2.1.0; R Foundation for
number of gradients with diffusion                           withdrawal of life support and may not         Statistical Computing, Vienna, Austria).
weighting (b value       1000 sec/mm2)                       have had an opportunity to show recov-         P .05 was considered to indicate a sig-
ranged from six to 25, with one to three                     ery as measured on the modified Rankin          nificant difference for all analyses.
images with a b value of 0 sec/mm2.                          scale. Thus, eye opening (spontaneous or
    Because DW imaging changes in co-                        in response to stimulation) within the 1st
matose patients with cardiac arrest are                      week was also used as a surrogate mea-          Results
often diffusely distributed and are lim-                     sure of potential good outcome. Differ-
ited to only qualitative evaluation and                      ences between patients with eye opening        Patient Demographic Data
are therefore difficult to compare across                     and those with no eye opening were com-        Table 1 shows demographic data for the
subjects, we examined whether the se-                        pared by using a two-tailed Wilcoxon           patients who exhibited early eye open-
verity of reduction in ADC can be used                       rank-sum test (JMP, version 7.0, SAS In-       ing (n 30) and those who did not (n
to objectively assess the extent of                          stitute, Cary, NC). No eye opening is a        50). Of the 80 patients, 64 died between

Radiology: Volume 252: Number 1—July 2009 ▪                                                                                 175
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                              Wu et al

1 and 22 days (median, 5 days) after             scores 3). Of the patients who under-          those not treated with hypothermia;
cardiac arrest. There were significant            went induced hypothermia (n           14),     this was likely due to the small number
differences (P .001) in 6-month mod-             seven (50%) opened their eyes (six pa-         of patients.
ified Rankin scale scores between pa-             tients by day 1 and one patient by day
tients with eye opening and patients             3), while five (36%) had a 6-month              Regional Analysis
without eye opening (Fig 2). Among the           modified Rankin scale score of 3 or less,       The putamen exhibited significantly lower
patients with eye opening, 17 opened             compared with 23 (35%) patients who            (P .001) median ADC (expressed as
their eyes on day 1 (57%), 10 on day 3           showed eye opening and nine (14%)              10 6 mm2/sec) (700 [interquartile range
(33%), and three on day 7 (10%). Of              who had a 6-month modified Rankin               {IQR}, 570 –770]), and the insula exhib-
these 30 patients, 15 died because of            scale score of 3 or less in the group that     ited significantly higher (P .01) median
withdrawal of care, one experienced an-          did not undergo hypothermia treatment          ADC (860 [IQR, 750 –920]), than the
other unexpected cardiac arrest and              (n 66). There were no significant dif-          other regions in the brain (cerebellum:
subsequently died, and 14 survived (13           ferences in terms of outcome between           770 [IQR, 690 – 810]; frontal lobe: 780
with 6-month modified Rankin scale                patients treated with hypothermia and          [IQR, 710 – 840]; occipital lobe: 760 [IQR,

 Figure 1

                                                                                               Figure 1: Images created from probabilistic atlas
                                                                                               show different color codes that represent the proba-
                                                                                               bility that tissue found at a position belongs to a given
                                                                                               tissue region (ie, white matter; caudate nucleus; puta-
                                                                                               men; thalamus; cerebellum; insula; frontal, occipital,
                                                                                               parietal, and temporal lobes) in (a) axial, (b) sagittal,
                                                                                               and (c) coronal views. For clarity, only tissue with a
                                                                                               probability of classification of greater than 50% is
                                                                                               shown. (Images generated by using Display; Mon-
                                                                                               treal Neurological Institute, Montreal, Quebec, Can-
                                                                                               ada.) (d, e) Coregistered axial (d) ADC and (e) DW
                                                                                               imaging maps in 67-year-old woman imaged 3 days
                                                                                               after cardiac arrest with a 6-month modified Rankin
                                                                                               scale score of 6.

176                                                                           ▪ Radiology: Volume 252: Number 1—July 2009
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                                                                    Wu et al

                                                               Table 1
600 – 820]; parietal lobe: 760 [IQR, 640 –
850]; temporal lobe: 820 [IQR, 760 –                             Imaging and Clinical Characteristics in Study Population
890]; caudate nucleus: 770 [IQR, 700 –
                                                                                                                     All Patients    Patients with Eye   Patients without Eye
840]; thalamus: 750 [IQR, 680 – 810]; and
                                                                 Parameter                                           (n 80)          Opening (n 30)      Opening (n 50)         P Value
white matter: 730 [IQR, 680 – 800]). Me-
dian whole-brain ADC in patients with                                No. of women*                                   31 (39)         10 (33)             21 (42)                  .5
eye opening was significantly higher than                             Age (y)
that in patients without eye opening                                    All patients                                 57 16           57 16               57 16                    .99
(Table 2). All regions, with the exception                              Women                                        57 17           58 18               56 17                    .7
of the thalamus, exhibited significantly                                 Men                                          57 16           56 16               58 16                    .7
more severe ADC reductions in patients                               No. of patients treated with hypothermia        14              7                   7                        .4
                                                                     Glasgow Coma Scale score at admission†           4 (3–8)         4 (3–8)             4 (3–7)                 .1
who did not experience eye opening. Pa-
                                                                     Time to MR imaging (d)†                          2 (0–10)        3 (0–10)            2 (0–6)                 .2
tients with poor outcome at 6 months
(modified Rankin scale score 3) exhib-                            Note.—Unless otherwise stated, data are means       standard deviations.
ited significantly lower ADCs in all re-                          *Data in parentheses are percentages.
gions (Table 3) with the exception of the                        †
                                                                     Data are medians, with ranges in parentheses.
thalamus, which showed marginally dif-
ferent ADC (P       .06). Among patients
who experienced early eye opening,
ADCs (expressed as 10 6 mm2/sec) in
the putamen (730 [IQR, 670 –750]) were                         Figure 2
significantly lower (P      .02) in patients
with poor outcome at 6 months (n 17)
than in patients with good outcome at 6
months (n 13) (760 [IQR, 720 – 840]).
For patients who underwent induced hy-
pothermia, the caudate nucleus (P .02)
and the putamen (P        .03) had signifi-
cantly higher ADCs, while the insula (P
.05) demonstrated a tendency toward a
higher ADC.

Early versus Late MR Imaging
There were significant differences in
whole-brain ADCs between patients
with and those without eye opening for
studies performed less than 3 days and
those performed 3 or more days after                           Figure 2: Graph shows distribution of 6-month modified Rankin scale (mRS) scores for patients with early
cardiac arrest (Table 4). For studies                          eye opening (n 30) and patients without early eye opening (n 50). The median value for both groups was
performed less than 3 days after the                           6; however, a significant difference (P .001) between the groups is evident. Among the patients character-
event, all regions except the cerebellum                       ized as having a good outcome according to a modified Rankin scale score of 3 or less (n 14), only one
showed significant differences between                          patient did not have eye opening within the 1st week after arrest.
eye-opening groups. For studies per-
formed 3 or more days after the event,
differences between patients with and                        .055). For studies performed less than 3                                   Patients who underwent induced
those without eye opening in terms of                        days after the event, the frontal, occipital,                          hypothermia tended to undergo imaging
ADCs for all regions were less pro-                          and parietal lobes, as well as the cau-                                at later time points (P   .001). There
nounced compared with the early imag-                        date nucleus and the putamen, exhib-                                   was no difference in ADCs between pa-
ing group except in the occipital lobe.                      ited significantly reduced ADCs. For                                    tients who underwent hypothermia
With outcome judged according to                             studies performed 3 or more days af-                                   (n 11) and those who did not (n 21)
6-month modified Rankin scale scores,                         ter the event, differences between pa-                                 imaged 3 or more days after the event.
whole-brain ADCs were significantly                           tients with good and those with poor                                   However, for patients imaged less than
lower in patients with poor outcomes                         modified Rankin scale scores became                                     3 days after cardiac arrest, there re-
(modified Rankin scale score         3) for                   more pronounced for all regions ex-                                    mained a tendency for patients who un-
late imaging studies but had a tendency to                   cept for the striatum, thalamus, and                                   derwent induced hypothermia (n      3)
be lower in patients imaged early (P                         cerebellum.                                                            to have larger ADCs in the caudate nu-

Radiology: Volume 252: Number 1—July 2009 ▪                                                                                                                        177
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                                                    Wu et al

cleus (P      .1), insula (P     .1), and                         those with poor (n    7) outcomes in                 icantly correlated with modified
putamen (P       .06) than patients who                           terms of modified Rankin scale score.                 Rankin scale score (Spearman
did not undergo hypothermia (n 45).                                                                                      0.36, P     .001, Fig 3). No patients
For the subset of patients with eye open-                         Logistic Regression Analysis                         with severe ( 665       10 6 mm2/sec)
ing and early imaging (n 15), no signif-                          In univariate analysis, considering                  reductions in median whole-brain
icant difference was found between those                          whole-brain median ADC, age, sex,                    ADC experienced either eye opening
with good (n       5) and poor (n      10)                        baseline Glasgow Coma Scale score,                   or good outcome according to the
outcomes according to the modified                                 whether hypothermia was induced,                     modified Rankin scale at 6 months.
Rankin scale; this was likely due to the                          and day of MR imaging, only whole-                   Multiple logistic regression analysis
small sample size. For patients with eye                          brain median ADC (P         .005) signifi-            incorporating all of the above param-
opening and late imaging (n 15), there                            cantly predicted 6-month modified                     eters revealed that whole-brain ADC
were significant differences in ADCs in                            Rankin scale score, although hypo-                   continued to be a significant factor
the whole brain (P .04), white matter                             thermia treatment (P       .06) and day              (P    .004). When the different brain
(P     .04), and putamen (P       .04) be-                        of MR imaging (P        .05) had a ten-              regions were included in the model
tween patients with good (n        8) and                         dency toward significance. ADC signif-                with whole-brain ADC by using back-
                                                                                                                       ward stepwise regression, only ADCs
                                                                                                                       in the occipital (P .006) and parietal
 Table 2                                                                                                               (P    .009) lobes and in the putamen
                                                                                                                       (P .04) were significant. The result-
   Median Whole-Brain and Regional ADCs in Patients with and Those without Eye
   Opening                                                                                                             ant model had a specificity of 100%
                                                                                                                       (95% confidence interval [CI]: 73%,
   Location                          Eye Opening (n     30)               No Eye Opening (n   50)   P Value            100%) and a sensitivity of 80% (95%
      Whole brain                    800 (740–850)                        730 (610–790)               .001             CI: 68%, 89%) compared with the uni-
      White matter                   770 (710–830)                        710 (580–770)               .002             variate model that incorporated only
      Caudate nucleus                790 (750–830)                        720 (670–840)               .03              whole-brain ADC, which had the same
      Putamen                        740 (700–790)                        650 (530–770)               .005             specificity but a sensitivity of 41%
      Thalamus                       780 (730–810)                        740 (650–790)               .09              (95% CI: 29%, 54%) (P        .001). For
      Cerebellum                     790 (740–820)                        740 (660–800)               .009             prediction of poor outcome by using
      Frontal lobe                   840 (780–870)                        750 (630–820)               .001             eye opening and whole-brain ADC,
      Insula                         910 (830–940)                        810 (680–890)               .001             specificity was 100% (95% CI: 86%,
      Occipital lobe                 810 (770–870)                        680 (530–790)               .001             100%) and sensitivity was 30% (95%
      Parietal lobe                  830 (760–890)                        690 (530–810)               .001             CI: 18%, 45%).
      Temporal lobe                  870 (810–910)                        780 (680–860)               .001
   Note.—Data are ADCs (   10       mm2/sec), with IQRs in parentheses.
                                                                                                                       Our results demonstrate that ADC maps
                                                                                                                       obtained during routine MR imaging may
 Table 3                                                                                                               be valuable in assessing the state of brain
                                                                                                                       damage and, hence, the potential for re-
   Median Whole-Brain and Regional ADCs versus Outcome according to 6-month                                            covery in patients who are comatose after
   Modified Rankin Scale Score                                                                                          cardiac arrest. The degree of ADC reduc-
                                      Modified Rankin Scale                  Modified Rankin Scale                       tion associated with poor outcome ap-
   Location                           Score 3 (n 14)                        Score 3 (n 66)          P Value            pears to depend on the region of tissue
                                                                                                                       and the timing of imaging. We found that
      Whole brain                     830 (740–900)                         750 (680–810)             .001
                                                                                                                       severe ADC depression within the first
      White matter                    790 (720–860)                         730 (620–780)             .005
                                                                                                                       few days of global anoxic brain injury is
      Caudate nucleus                 800 (780–900)                         750 (680–830)             .01
      Putamen                         760 (720–850)                         680 (540–760)             .001
                                                                                                                       highly specific for permanent injury. We
      Thalamus                        790 (720–850)                         750 (660–790)             .06              found a strong association between the
      Cerebellum                      810 (770–830)                         750 (680–800)             .007             degree and location of ADC changes and
      Frontal lobe                    840 (770–900)                         760 (680–830)             .003             different outcome measures, including
      Insula                          920 (840–990)                         830 (730–910)             .005             eye opening and score on the modified
      Occipital lobe                  850 (770–910)                         730 (570–810)             .001             Rankin scale, a well-validated scale of
      Parietal lobe                   850 (740–910)                         750 (570–830)             .003             neurologic disability. At imaging per-
      Temporal lobe                   880 (810–940)                         800 (740–880)             .009             formed within the first 3 days after car-
                                                                                                                       diac arrest, ADC differences involved
   Note.—Data are ADCs (   10       mm2/sec), with IQRs in parentheses.
                                                                                                                       mostly the cortical regions and striatum,

178                                                                                                  ▪ Radiology: Volume 252: Number 1—July 2009
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                                                                   Wu et al

                                                               Table 4
whereas in later studies, the changes
were restricted to cortical regions and                          P Values for Difference in Median ADCs between Patients with Good and Those with
subcortical white matter. Changes in the                         Poor Outcome Dichotomized by Whether MR Imaging Was Performed Early or Late
cerebellum were more evident in patients
                                                                                                                                                    P Value for Modified Rankin Scale
undergoing later imaging.                                                                       P Value for Eye Opening vs No Eye                  Score 3 vs Modified Rankin Scale
     We speculate that the lack of differ-                                                                   Opening                                            Score 3
ences in the later time points in the stri-                                              Early MR              Late MR                           Early MR            Late MR
atum and thalamus could be due to                                Location                Imaging (n 15) Imaging (n 15)                           Imaging (n 6) Imaging (n 8)
pseudonormalization. Animal models
have shown that with the restoratron of                              Whole brain           .001                    .05                           .06                   .007
perfusion, brain tissue experiences rapid                            White matter          .003                    .1                            .08                   .01
pseudonormalization (12), only to deteri-                            Caudate nucleus       .01                     .9                            .04                   .3
                                                                     Putamen               .03                     .2                            .01                   .03
orate again at subsequent time points
                                                                     Thalamus              .04                     .9                            .1                    .5
(13). We hypothesize that a similar pat-
                                                                     Cerebellum            .08                     .08                           .07                   .08
tern may occur in patients resuscitated
                                                                     Frontal lobe          .001                    .06                           .04                   .03
after cardiac arrest. The observed time
                                                                     Insula                .009                    .03                           .07                   .04
course of brain damage that our results                              Occipital lobe        .003                    .001                          .02                   .001
suggest—that is, initial ADC reduction in                            Parietal lobe         .001                    .01                           .04                   .01
the striatum and thalamus, followed by                               Temporal lobe         .002                    .1                            .08                   .06
the cortex and then the subcortical white
matter—may be an indication of ongoing                           Note.—Early MR imaging ( 3 days after cardiac arrest) was performed in 48 patients; late MR imaging ( 3 days after cardiac
                                                                 arrest) was performed in 32 patients. In parentheses are the numbers of patients in each group with good outcome (eye
tissue damage as a result of secondary                           opening or modified Rankin scale score 3).
energy failure (27) and apoptotic pro-
cesses (28). We suspect that the rate at
which an individual patient reaches these
stages will depend on the time to resusci-                     Figure 3
tation, the etiology of the cardiac arrest
(29), and the patient’s comorbidities. A
prospective study involving longer obser-
vational times for neurologic recovery
and predefined imaging time points is
clearly needed to better characterize the
dynamics of brain damage in these pa-
tients. If a pattern representing a high
likelihood of poor outcome could be de-
termined, the early withdrawal of care
from patients who would otherwise have
a chance for recovery could be avoided.
Alternatively, a pattern that demon-
strates whether salvageable tissue still ex-
ists may be identified, and therefore a
patient could undergo therapy such as
                                                               Figure 3: Scatterplot of ADC versus modified Rankin scale (mRS) score as a function of time of MR imag-
neuroprotective treatment or at least con-
                                                               ing (early, 3 days; late, 3 days) and whether or not the patient experienced eye opening. Patients with
tinued supportive care. Complementary
                                                               whole-brain ADCs less than 665 10 6 mm2/sec had poor outcomes, defined as no eye opening or a
to determining whether therapy can be
                                                               6-month modified Rankin scale score of 6, regardless of the timing of the MR imaging study.
administered successfully, our findings
also suggest that ADC may be helpful in
monitoring the benefits of therapy, as                        tients with severe ( 665 10 6 mm2/                                 brain metrics, calculating medi-
shown by the higher measured ADCs in                         sec) whole-brain median ADC reduc-                                 an ADC automatically becomes fairly
patients who underwent induced hypo-                         tions experienced either eye opening                               straightforward when one is assessing
thermia, a finding consistent with findings                    or good outcome according to the                                   the likelihood of recovery. By per-
in animal studies (30,31).                                   modified Rankin scale at 6 months.                                  forming multivariate analysis, we
     As a preliminary attempt to iden-                       This suggests that marked whole-                                   were able to increase the sensitivity of
tify patterns with poor outcome, we                          brain median ADC reductions may be                                 MR imaging– based metrics (from
related whole-brain median ADC to                            a reflection of severe, irreversible                                41% to 80%) without reducing speci-
clinical outcome. In our cohort, no pa-                      brain damage. By relying on whole-                                 ficity (100%). However, a comparison

Radiology: Volume 252: Number 1—July 2009 ▪                                                                                                                        179
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                              Wu et al

of these results with current clinical           sion, but its value is limited because of       and may therefore be a valuable tool for
reference standards, such as somato-             the small number of patients studied.           clinical assessment. The location and de-
sensory evoked potentials (3), should                 The timing of MR imaging has also          gree of ADC depression is associated with
be performed, as well as validated               previously been found to be a potential         clinical outcome measures, both for imag-
against an independent cohort of pa-             confounding factor, as is consistent with       ing performed acutely and that per-
tients.                                          our findings. Els et al (33) evaluated 12        formed subacutely after arrest. Future
    Prior studies of MR imaging in car-          patients with DW imaging within 36              studies are needed to prospectively eval-
diac arrest have been limited by small           hours of a cardiac arrest lasting 10 –25        uate these findings and will hopefully
sample sizes and an absence of quanti-           minutes. There was an attempt in eight of       prove MR imaging to be a highly sensitive
tative data. Wijdicks et al (14) evalu-          the 12 patients to quantify the degree of       and specific diagnostic tool in prognosti-
ated 10 comatose survivors of cardiac            ADC depression. Patients who had short          cating recovery from coma after cardiac
arrest with DW imaging, limiting their           resuscitation times and limited DW imag-        arrest.
study to only those patients who had             ing changes went on to make a good re-
                                                                                                 Acknowledgments: The authors acknowledge
preserved pupillary reflexes and normal           covery, whereas the others progressed to
                                                                                                 Mark Vangel, PhD, for manuscript review and
somatosensory evoked potentials. Pa-             a minimally conscious or persistent vege-       statistical guidance.
tients underwent MR imaging per-                 tative state. Others have cautioned that a
formed at variable time points within 15         normal MR imaging study this early can
days of cardiac arrest. Eight patients           lead to falsely optimistic results and that
                                                                                                  1. Thel MC, O’Connor CM. Cardiopulmonary
with diffuse signal intensity abnormali-         changes consistent with a more severe
                                                                                                     resuscitation: historical perspective to re-
ties had no recovery beyond a severely           injury may be seen on subsequent imag-              cent investigations. Am Heart J 1999;137:
disabled state, whereas two patients             ing studies (13,34–36).                             39 – 48.
with limited or no changes later awak-                Our study was limited by being a ret-
                                                                                                  2. Pusswald G, Fertl E, Faltl M, Auff E. Neuro-
ened. Arbelaez et al (15) performed a            rospective analysis with selection bias.            logical rehabilitation of severely disabled
retrospective analysis of 11 MR imaging          We did not control whether a patient un-            cardiac arrest survivors. II. Life situation of
studies in 10 patients with prolonged            derwent MR imaging or what the timing               patients and families after treatment. Resus-
( 6 minutes) cardiac arrest. They sep-           of the MR imaging examination would be.             citation 2000;47:241–248.
arated their findings on the basis of the         MR imaging may not be appropriate for            3. Levy DE, Caronna JJ, Singer BH, Lapinski
timing of the imaging examination after          all comatose patients with cardiac arrest,          RH, Frydman H, Plum F. Predicting outcome
the arrest: acute ( 24 hours), early             such as those with ongoing hemodynamic              from hypoxic-ischemic coma. JAMA 1985;
subacute (24 hours to 13 days), late             instability, a poor long-term prognosis on          253:1420 –1426.
subacute (14 –20 days), and chronic              the basis of other comorbidities, or a con-      4. Edgren E, Hedstrand U, Nordin M, Rydin E,
( 21 days). They suggested that there            traindication to MR imaging (eg, pace-              Ronquist G. Prediction of outcome after car-
is an evolution in DW imaging signal             maker or automatic implantable cardio-              diac arrest. Crit Care Med 1987;15:820 –
intensity abnormalities from the basal           verter-defibrillator). We could not dictate          825.
ganglia and cerebellum (acute), to the           whether a patient was allowed to survive         5. Zandbergen EG, Hijdra A, Koelman JH, et al.
cortex and basal ganglia (early sub-             for a longer period of time to determine if         Prediction of poor outcome within the first 3
acute), to the white matter (late sub-           he or she would recover; thus, many pa-             days of postanoxic coma. Neurology 2006;66:
acute). However, the number of pa-               tients may have had a modified Rankin                62– 68.
tients in that study was small, and only         scale score of 6 because of withdrawal of        6. Fogel W, Krieger D, Veith M, et al. Serum
one patient underwent imaging twice,             life-sustaining therapy. For this reason,           neuron-specific enolase as early predictor of
making it difficult to know whether               we also assessed the outcome measure of             outcome after cardiac arrest. Crit Care Med
these changes were evolutionary or rep-          early eye opening as an indication of pa-
resentative of individual patient charac-        tients who may have been showing some            7. Prohl J, Rother J, Kluge S, et al. Prediction
teristics. Chalela et al (16) retrospec-         recovery prior to withdrawal of care. This          of short-term and long-term outcomes after
tively analyzed MR imaging studies in            self-fulfilling prophecy is a common prob-           cardiac arrest: a prospective multivariate ap-
                                                                                                     proach combining biochemical, clinical, elec-
seven patients who underwent imaging             lem in many studies of patients with car-
                                                                                                     trophysiological, and neuropsychological in-
within 7 days of cardiac arrest. They            diac arrest. Investigators in future studies        vestigations. Crit Care Med 2007;35:1230 –
described a pattern of signal intensity          may wish to pursue a longer observation             1237.
abnormalities that primarily affected            time or other potential outcome mea-
                                                                                                  8. Mullie A, Verstringe P, Buylaert W, et al.
the white matter, consistent with a po-          sures that may be used while the patient            Predictive value of Glasgow coma score for
tential selective myelinolytic process           is still alive, such as eye opening or a            awakening after out-of-hospital cardiac ar-
triggered by anoxia followed by ische-           positive change in the Glasgow Coma                 rest. Cerebral Resuscitation Study Group of
mia, such as with a respiratory arrest           Scale score.                                        the Belgian Society for Intensive Care. Lan-
leading to a cardiac arrest (32). The                 In summary, severe ADC reduction is            cet 1988;1:137–140.
study of Chalela et al also provided a           a strong prognosticator of poor outcome          9. Wijdicks EF, Hijdra A, Young GB, Bassetti
quantitative analysis of ADC depres-             in comatose patients with cardiac arrest            CL, Wiebe S. Practice parameter: predic-

180                                                                            ▪ Radiology: Volume 252: Number 1—July 2009
NEURORADIOLOGY: MR Imaging for Outcome Prediction in Cardiac Arrest                                                                                         Wu et al

    tion of outcome in comatose survivors af-                        ing MR imaging. Radiology 1999;212:785–          neurological outcomes after a short period of
    ter cardiopulmonary resuscitation (an evi-                       792.                                             focal brain ischemia in rats. Ann Neurol
    dence-based review): report of the Quality                                                                        2000;48:236 –244.
                                                             18. Reese TG, Heid O, Weisskoff RM, Wedeen
    Standards Subcommittee of the American                                                                        28. Li F, Silva MD, Sotak CH, Fisher M. Tempo-
                                                                 VJ. Reduction of eddy-current-induced dis-
    Academy of Neurology. Neurology 2006;                                                                             ral evolution of ischemic injury evaluated
                                                                 tortion in diffusion MRI using a twice-refo-
    67:203–210.                                                                                                       with diffusion-, perfusion-, and T2-weighted
                                                                 cused spin echo. Magn Reson Med 2003;49:
                                                                 177–182.                                             MRI. Neurology 2000;54:689 – 696.
10. Yanagawa Y, Un-no Y, Sakamoto T, Okada
    Y. Cerebral density on CT immediately after                                                                   29. Singhal AB, Topcuoglu MA, Koroshetz WJ.
                                                             19. Collins DL, Neelin P, Peters TM, Evans AC.
    a successful resuscitation of cardiopulmo-                                                                        Diffusion MRI in three types of anoxic en-
                                                                 Automatic 3D intersubject registration of
    nary arrest correlates with outcome. Resus-                                                                       cephalopathy. J Neurol Sci 2002;196:37– 40.
                                                                 MR volumetric data in standardized Ta-
    citation 2005;64:97–101.                                                                                      30. Yenari MA, Onley D, Hedehus M, et al. Diffu-
                                                                 lairach space. J Comput Assist Tomogr
                                                                 1994;18:192–205.                                     sion- and perfusion-weighted magnetic reso-
11. Hjort N, Christensen S, Solling C, et al. Ische-
                                                                                                                      nance imaging of focal cerebral ischemia and
    mic injury detected by diffusion imaging 11              20. UCLA Laboratory of Neuro Imaging. ICBM-              cortical spreading depression under conditions
    minutes after stroke. Ann Neurol 2005;58:                    452 T1 5th order polynomial warps atlas.             of mild hypothermia. Brain Res 2000;885:
    462– 465.                                                    Los Angeles, Calif: UCLA Laboratory of               208 –219.
                                                                 Neuro Imaging, 2008.
12. Pierpaoli C, Alger JR, Righini A, et al. High                                                                 31. Piepgras A, Elste V, Frietsch T, Schmiedek
    temporal resolution diffusion MRI of global              21. UCLA Laboratory of Neuro Imaging. ICBM               P, Reith W, Schilling L. Effect of moderate
    cerebral ischemia and reperfusion. J Cereb                   probabilistic atlases. Los Angeles, Calif:           hypothermia on experimental severe sub-
    Blood Flow Metab 1996;16:892–905.                            UCLA Laboratory of Neuro Imaging, 2008.              arachnoid hemorrhage, as evaluated by ap-
                                                                                                                      parent diffusion coefficient changes. Neu-
13. Dijkhuizen RM, Knollema S, van der Worp                  22. Mazziotta J, Toga A, Evans A, et al. A prob-         rosurgery 2001;48:1128 –1134.
    HB, et al. Dynamics of cerebral tissue injury                abilistic atlas and reference system for the
    and perfusion after temporary hypoxia-ische-                 human brain: International Consortium for        32. Ginsberg MD, Hedley-Whyte ET, Richardson
    mia in the rat: evidence for region-specific                  Brain Mapping (ICBM). Philos Trans R Soc             EP Jr. Hypoxic-ischemic leukoencephalopathy
    sensitivity and delayed damage. Stroke                       Lond B Biol Sci 2001;356:1293–1322.                  in man. Arch Neurol 1976;33:5–14.
    1998;29:695–704.                                                                                              33. Els T, Kassubek J, Kubalek R, Klisch J. Dif-
                                                             23. Pierpaoli C, Jezzard P, Basser PJ, Barnett A.        fusion-weighted MRI during early global ce-
14. Wijdicks EF, Campeau NG, Miller GM. MR                       Diffusion tensor MR imaging of the human             rebral hypoxia: a predictor for clinical out-
    imaging in comatose survivors of cardiac re-                 brain. Radiology 1996;201:637– 648.                  come? Acta Neurol Scand 2004;110:361–
    suscitation. AJNR Am J Neuroradiol 2001;                                                                          367.
                                                             24. Lowry R. VassarStats: Web site for statisti-
                                                                 cal computation.       34. Greer DM. MRI in anoxic brain injury. Neu-
15. Arbelaez A, Castillo M, Mukherji SK. Diffu-                  /lowry/VassarStats.html. Accessed January            rocrit Care 2004;1:213–215.
    sion-weighted MR imaging of global cerebral                  14, 2009.                                        35. Hald JK, Brunberg JA, Dublin AB, Wootton-
    anoxia. AJNR Am J Neuroradiol 1999;20:                   25. Newcombe RG. Two-sided confidence inter-              Gorges SL. Delayed diffusion-weighted MR
    999 –1007.                                                   vals for the single proportion: comparison of        abnormality in a patient with an extensive
                                                                 seven methods. Stat Med 1998;17:857– 872.            acute cerebral hypoxic injury. Acta Radiol
16. Chalela JA, Wolf RL, Maldjian JA, Kasner
    SE. MRI identification of early white matter              26. Beck JR, Shultz EK. The use of relative oper-
    injury in anoxic-ischemic encephalopathy.                                                                     36. Hossmann KA, Fischer M, Bockhorst K,
                                                                 ating characteristic (ROC) curves in test per-
    Neurology 2001;56:481– 485.                                                                                       Hoehn-Berlage M. NMR imaging of the appar-
                                                                 formance evaluation. Arch Pathol Lab Med
                                                                                                                      ent diffusion coefficient (ADC) for the evalua-
17. Sorensen AG, Wu O, Copen WA, et al. Hu-                                                                           tion of metabolic suppression and recovery af-
    man acute cerebral ischemia: detection of                27. Li F, Silva MD, Liu KF, et al. Secondary             ter prolonged cerebral ischemia. J Cereb
    changes in water diffusion anisotropy by us-                 decline in apparent diffusion coefficient and         Blood Flow Metab 1994;14:723–731.

Radiology: Volume 252: Number 1—July 2009 ▪                                                                                              181

Shared By: