spiral.imperial.ac.uk · Web viewWe considered English and non-English articles and original...
Transcript of spiral.imperial.ac.uk · Web viewWe considered English and non-English articles and original...
1
Cover title: AF detection after TIA: systematic review/meta-analysis
Title: Cardiac monitoring for detection of atrial fibrillation after TIA: a systematic review and meta-analysis
Eleni Korompoki1 MD PhD, Angela Del Giudice1 MD, Steffi Hillmann2 MPH, Uwe Malzahn2,3 PhD, David J Gladstone4 MD, PhD, Peter Heuschmann2,3 MD, MPH, Roland Veltkamp1 MD
1 Division of Brain Sciences, Department of Stroke Medicine, Imperial College, London, UK
2 Institute of Clinical Epidemiology and Biometry, University of Würzburg, Germany
3 Clinical Trial Center , Comprehensive Heart Failure Center,UniversityHospital Würzburg, Germany
4 Division of Neurology, Department of Medicine, University Toronto, Ontario, Canada
Corresponding author:
Eleni Korompoki, MD PhD
Department of Stroke Medicine
Imperial College London
Fulham Palace Road
London, W6 8RF, UK
Phone +44-20-33130133
Supplemental material (tables I-V, e-references, supplementary figure)
Tables: 2
Figures: 5
References: 55
Abstract: 245 words
Manuscript: 5141 words
Key words: ischemic attack, transient; atrial fibrillation; electrophysiology
2
Abstract
Background and Purpose: The detection rate of atrial fibrillation (AF) has not been studied
specifically in TIA patients although extrapolation from findings in ischemic stroke (IS) may
be inadequate. We conducted a systematic review and meta-analysis to determine the rate of
newly diagnosed AF using different methods of ECG monitoring in TIA.
Methods: A comprehensive literature search was performed following a pre-specified
protocol in accordance with the PRISMA statement. Prospective observational studies and
randomized controlled trials were considered that included TIA patients who underwent
cardiac monitoring for >12 hours. Primary outcome was frequency of detection of AF lasting
at least 30 sec. Analyses of subgroups and of duration and type of monitoring were
performed.
Results: 17 studies enrolling 1163 patients were included. Overall, the pooled AF detection
rate was 4% (95% CI: 2%-7%). Yield of monitoring was higher in selected (in terms of age,
pre-screening for arrhythmias and cause of TIA) than in unselected cohorts (7% vs 3%).
Pooled mean AF detection rates rose with duration of monitoring: 4% (24 hours), 5% (24
hours to7 days) and 6% (>7 days), respectively. The yield of non-invasive monitoring was
significantly lower than that of invasive monitoring (4% vs. 11%). Significant heterogeneity
was observed among studies (I2=60.61%).
Conclusion: This first meta-analysis of AF detection in TIA patients suggests that the rate of
AF detection is lower in TIA patients than reported for IS and TIA cohorts in previous meta-
analyses. Prospective studies are needed to determine the optimal diagnostic procedure for
AF detection in TIA.
3
Introduction
Sensitive diagnosing of atrial fibrillation (AF) represents a major objective in stroke and TIA
patients (1). While growing evidence guides diagnostic work-up for detection of AF in stroke
patients, the yield of various modalities of cardiac monitoring in TIA patients has not been
studied specifically (2, 3). Extrapolation of findings in stroke cohorts to TIA patients may be
inadequate. Although TIAs are considered as part of the spectrum of ischemic stroke (IS) (4),
the diagnosis of TIA is often based on medical history alone and is less definite. Moreover,
the setting of cardiac diagnostic work-up for strokes and TIAs differs substantially. In many
health care systems, TIAs are evaluated as outpatients (5-7) whereas strokes are admitted to
monitored stroke units (8, 9), Even if TIA patients are hospitalized, length of monitoring is
shorter (10). No systematic review assessing different techniques for AF detection in TIA
patients is available. Thus, the optimal duration, method and timing of monitoring and the
best candidates for prolonged monitoring in TIA are ill defined.
We performed a systematic review and meta-analysis to determine the overall rate of newly
detected AF after TIA. We also explored detection rates in selected (in terms of age, extent of
cardiac tests before enrollment, cryptogenic cause) and unselected TIA cohorts, and
evaluated the impact of duration and type of monitoring.
Methods
Search strategy and data extraction
We performed a systematic review following PRISMA guidance (figure 1)(11) and a pre-
specified study protocol. Relevant studies were identified using thesaurus subject headings
(MEDLINE and EMBASE) and free text terms (table s-1). Google scholar, Cochrane library,
clinical trial registries and references lists from all included articles and abstracts were also
4
assessed. Pre-defined eligibility criteria regarding study design, duration of monitoring, AF
definition were applied by two independent investigators (EK, SH). If necessary authors
were asked to provide information on eligibility criteria (e.g. study design, AF definition,
exclusion of known AF cases, duplicate publications). Corresponding authors of studies
including both IS and TIA patients were contacted by email to provide data about
characteristics and AF detection rates in TIA patients specifically. For studies which
examined more than one method of ECG monitoring in the same cohort, detection rates were
calculated for each method. Consequently, the number of pooled estimates exceeded the
number of eligible studies.
Eligibility criteria
We considered English and non-English articles and original abstracts published between
1966 and March 1, 2015. We included studies that enrolled TIA patients or mixed
populations of TIA and IS patients who underwent ECG monitoring for at least 12 hours.
Only randomized controlled trials and prospective observational studies reporting AF
episodes lasting ≥ 30 sec were included. Studies including patients with known AF or with
AF diagnosed during screening were excluded.
Outcome measures
The main outcome measure was the overall rate of newly detected AF. The proportion of
patients with AF in each subgroup was calculated by dividing the number of patients with
newly detected AF by the total number of patients included in this group.
Three subgroup analyses were performed. First, we determined the AF detection rate in
unselected versus selected patients. Included studies used various selection criteria. Some
included only “cryptogenic” TIAs or only supposedly embolic TIAs. Other selection criteria
5
related to duration and methods of screening for cardiac arrhythmias before enrolment. Some
studies only enrolled patients above a defined age. Second, we reported AF detection rates
depending on duration of ECG monitoring which was categorized into intervals ≤ 24 hours,
between 24 hours and 7 days, and > 7 days. Finally, we estimated the AF detection rate
depending on the non-invasive versus invasive monitoring).
Risk of bias assessment
We assessed the quality of the studies according to the Cochrane handbook (12). We used a
funnel plot and Egger regression test in order to assess publication bias (13). For the funnel
plot we plotted the AF detection rate (based on the double arcsine transformation) against the
precision (1/standard error). A p value <0.1 was considered significant for publication bias.
Statistical analysis
Four meta-analyses were performed to determine the overall AF detection rate in the total
population and in predefined subgroups. We used the random-effects model for all analyses
because a high degree of heterogeneity was expected among the included cohorts.
Confidence intervals (CI) were calculated by the exact binomial method. Freeman-Tukey
double arcsine transformation was used to stabilize the variance. Heterogeneity was assessed
by the χ2-based Q statistics, a measure of variance on a relative scale, and by Ι2, a quantitative
measure of inconsistency, which is neither sensitive to the metric of the analyzed quantity nor
to the number of studies.Within a random effects model, heterogeneity between study
subgroups was assessed using Q-statistics (expressed as “Heterogeneity between groups” in
forest plots). Higher values of I2 or values of p<0.1 for the Q-statistic based χ2-test suggest
significant heterogeneity. Pooled proportion meta-analysis was performed using STATA
(version 13.0) (14).
6
Results
Twenty four articles, 16 full manuscripts and 8 abstracts, of 1205 identified citations fulfilled
the eligibility criteria (figure 1) (3, 15-37). Three studies (13%) were randomized controlled
trials and 21 (87%) were prospective observational (table s-2). After 7 studies were excluded
from the analysis because of missing data, 17 studies were included in the meta-analysis.
Therein, 25 monitored study subgroups were reported because some studies investigated
various diagnostic procedures in the same cohort (figure 2). Baseline characteristics were
available for 14 studies (table s-2). Only one study was performed exclusively in TIA patients
(32). We excluded 147 patients because in 17 cases AF was detected at presentation (27, 37),
and for 130 patients the detection rate was not available (15, 18, 25, 30, 34-36). Thus, a total
of 1163 patients from 17 studies were eligible for analysis (table s-2). Since 7 of these studies
investigated more than one modality in the same study cohort, the number of included study
subgroups exceeded the number of eligible studies. Thus, 1423 data sets from 25 cohorts
were included in our meta-analysis (3, 16, 17, 19-24, 26-29, 31-33, 37).
Overall AF detection rate
The overall pooled proportion of TIA patients with newly detected AF was 4% (95% CI: 2%-
7%) (figure 2). The study reporting the highest yield for AF detection (27%, 95% CI: 8%-
55%) used inpatient ECG monitoring (29). No AF episodes were detected in seven other
cohorts using different modalities (19, 20, 22, 26, 29, 31, 33). Significant heterogeneity was
observed among studies (Ι2=60.61%, p=0.00) (figure 2).
Subgroup analyses
Selected versus unselected cohorts
7
Data regarding unselected and selected cohorts are presented in tables 1 and 2, respectively.
Selection criteria differed substantially among groups (table 2, table s-4). AF was more likely
to be uncovered in selected (7%, 95% CI: 2%-14%) compared to unselected patients (3%,
95% CI: 2%-5%) with significant heterogeneity between groups (p=0.041) (figure 3).
Duration of monitoring
Extending the duration of monitoring from ≤24 hours to 7 days conferred a 1% higher
detection rate (4%; CI: 1%-8% vs. 5%; 95% CI: 2%-9%). Further extension of monitoring
beyond 7 days resulted in another 1% increase of AF detection (6%, 95% CI: 1%-12%). No
significant heterogeneity was observed among groups (p=0.558) (figure 4).
Invasive versus Non-invasive monitoring
Invasive methods yielded a higher detection rate than non-invasive modalities (11%: 95% CI:
2%-24% vs. 4%, 95% CI: 2%-7%). Significant heterogeneity existed between groups
(p=0.05). The heterogeneity among studies using invasive methods of monitoring was 0%
(figure 5) but the small number of studies precluded an estimation of between-study variance
with acceptable precision. Moreover, I2 is truncated to zero when the Q statistic is smaller
than its degrees of freedom (38).
Assessment of publication bias
The funnel plot showed asymmetry, suggesting a possible publication bias. We acknowledge
that the interpretation of funnel plot asymmetry may be inaccurate for assessment of
publication bias in meta-analyses of proportions, particularly in case of low proportions (39).
In this setting, the choice for both vertical and horizontal axes is challenging, and it can lead
to misinterpretation of funnel plot asymmetry (40). To quantitatively assess the risk of bias,
we performed an Egger regression test which was not significant (p=0.552), thus not allowing
8
to reject the null hypothesis that there is no small study effect (figure s-1). Overall 82% of the
included studies had possible selection bias. Potential funding bias was found in 29% of the
studies. We anticipate lower rate of reporting bias as in the vast majority (70% of included
studies) AF detection rates were provided directly by the corresponding authors after personal
communication. Other risk of bias are presented in table s-5.
Discussion
The major findings of this study are: (1) The overall rate of newly detected AF in TIA
patients is 4%. (2) The AF detection rate was 7% in selected compared to 3% in unselected
cohorts. (3) Invasive monitoring yielded substantially higher rates of AF detection than non-
invasive methods. (4) The yield of monitoring for > 7d confers limited added value compared
to 24 hours. (5) Inpatient ECG monitoring resulted in a relatively high rate of AF detection.
The overall new AF detection rate in TIA patients of 4% in our review is substantially lower
than rates reported in previous reviews of studies including both IS and TIA patients (i.e. 6.3-
11.5%) (41-43). Previous meta-analyses revealed considerably higher heterogeneity assessed
by I2 (88.2to 90.12%) compared to our study (I2=60.61%) (41, 43). A potential explanation
for both lower AF detection rates and lower heterogeneity in our meta-analysis could be that
we used stricter inclusion criteria. TIA patients only represent a relatively small proportion of
patients enrolled in studies included in previous meta-analyses on both IS and TIA patients.
Differences of baseline characteristics between IS and TIA patients are another potential
cause of different detection rates. Moreover, the variability in TIA definitions across the
included studies may have led to an overestimation of the total number of TIA patients (table
s-3). For example, brain imaging was not required to rule out conditions mimicking TIAs in
some studies, and TIA mimics are not at the same risk for subsequent stroke as patients with
a confirmed diagnosis of TIA (44). We considered only studies with AF episodes lasting
9
more than 30 sec. In contrast, previous reviews including mixed IS and TIA cohorts
attributed part of the observed heterogeneity and bias to less strict AF definitions (41, 42). An
alternative explanation for the lower prevalence of AF in TIA patients is that a proportion of
newly detected AF in IS is attributed to the infarct (i.e. “neurogenic AF”). As the majority of
TIA patients have no or only small infarcts (45) AF is unlikely to result from TIA. Among
studies included into our review, only the CRYSTAL-AF trial included exclusively patients
with infarcts on brain imaging (46) whereas the remaining studies recruited patients based on
symptom-duration alone (4) (table s-4).
The average incidence of AF was more than twice as high in selected compared to unselected
TIA patients. The wide range of detection rates (0%-27%) and the higher heterogeneity in
selected cohorts may arise from selection criteria applied in this group. Patients were
recruited into specific studies based on age, availability of cardiac tests before screening for
AF, and presumed or “cryptogenic” etiology of TIA. Even in the “cryptogenic TIA” group
inclusion criteria varied from just negative ECG and 24-hour Holter to pre-specified
extensive diagnostic work-up before enrollment (table s-4). Moreover, the time interval
between TIA and initiation of monitoring during which the investigations were performed
differed substantially among studies. In eight studies TIA patients received prolonged
monitoring within the first 72 hours after the index event (3, 20, 21, 24, 27-29, 37) whereas
other studies recruited patients later than 30 days after TIA (16, 17, 19, 22, 23, 31). In some
studies monitoring started immediately after admission in an inpatient setting (3, 20, 21, 24,
27-29, 33, 36, 37) whereas patients received a monitor as outpatients in other studies (16-19,
22, 23, 25, 26, 31, 32, 34, 35). Nevertheless, AF detection is more likely in selected patients
with suspected embolism compared to unselected populations (23, 31, 47). Older age, certain
clinical and neuroimaging findings (e.g. cortical deficits and embolic lesion pattern,
respectively) (48-50), and cardiovascular characteristics (e.g. higher risk stratification scores,
10
atrial premature beats, enlarged left atrium) (51, 52) may identify patients at higher risk of
AF. Stratification according to risk may increase the yield of extensive diagnostic
investigation.
Our finding that invasive monitoring yields much higher detection rates than non-invasive
techniques (figure 4) should be interpreted with caution. First, the number of TIA patients
who received an insertable device was small (44 patients), and it is likely that TIA patients
receiving invasive monitoring were selected (table s-3). Two of the three studies assessing
invasive monitoring reported an AF detection rate of 15% (17, 31) whereas the third study
failed to uncover any AF but only 3 young TIA patients were included (19). Importantly, our
comparison between invasive and non-invasive methods of monitoring is limited because the
non-invasive group comprised a wide variety of monitoring periods (i.e. 24 hours to 30 days).
Because the importance of the duration of monitoring for AF detection has been
unequivocally shown in stroke patients, we performed a separate meta-analysis addressing of
the impact of different monitoring periods. This analysis yielded only a limit additional yield
of monitoring beyond 7 days. It remains to be shown whether very extensive non-invasive
recording as performed over 30 days in the EMBRACE trial in stroke and TIA patients
results in a substantially higher rate of AF detection in unselected TIA patients (23).
In unselected and selected TIA cohorts inpatient ECG monitoring yielded the highest rate of
AF detection (3, 29). Rapid cardiac assessment of TIA patients is more likely to occur in TIA
inpatients with a recent episode. Recently revised guidelines recommend prolonged
monitoring for AF after stroke or TIA if no cause is apparent (7, 47). However, the optimal
setting, method, duration and timing of monitoring in TIA patients remain unknown as
prospective studies specifically addressing TIA patients are lacking.
11
Our study has strengths and limitations. We reported an overall AF detection rate of 4%
however our results should be interpreted with caution because of significant heterogeneity
between studies: We combined studies with different designs, methods, modalities, duration
of monitoring and cohort characteristics. However we performed three different subgroup
analyses in order to evaluate the detection rates in selected and unselected cohorts as well as
to explore the influence of the duration and modality (invasive versus non invasive) on AF
detection rates after TIA. We performed a comprehensive search of the literature and
followed a strict search protocol consistent with PRISMA guidance. On the other hand, we
obtained data from only 17 of 24 eligible studies which may have led to a selection bias.
However, the included studies reported on 89% of patients enrolled in all studies. Studies
which examined more than one modality for detecting AF in the same patient population (3,
20, 29) were considered as different independent studies in our meta-analysis as in previous
meta-analyses on the topic (41, 43). This may have resulted in a spuriously small estimate of
the variance of the effect size and in a high likelihood of a type I error. As sensitivity
analyses including only one modality for each study at a time provided similar results (data
not shown), we did not apply more complex methods of dealing with dependency such as
multi-level meta-analysis and robust variance estimation which also have limitations (53, 54).
Furthermore, we were unable to assess the interval between the index event and the initiation
of monitoring. However, our meta-analysis suggests that inpatient monitoring has a higher
yield of AF detection in both selected and unselected patients. Funnel plot showed
asymmetry suggesting a risk of bias. However, we tried to minimize publication bias by
searching all potential studies independently of the language and type of publication in
different databases. Finally, because most studies did not provide baseline characteristics for
risk stratification including age, gender and ABCD2 score further subgroup analyses were not
possible.
12
Summary/Conclusions
This systematic review and meta-analysis finds a substantially lower detection rate of AF in
TIA patients compared to previous analyses of mixed cohorts of predominantly stroke and
less TIA patients. Additional analyses suggest that selected patients or those undergoing
invasive monitoring have higher rates of AF detection than unselected patients. As
extrapolation from findings in stroke cohorts to TIA patients may be inadequate, prospective
studies are needed to determine the most appropriate method and duration of prolonged
cardiac monitoring in TIA patients specifically.
Acknowledgements
We are grateful to Dr Dominique Babuty, Dr Luisa Christensen, Dr Exuperio Díez Tejedor,
Dr Catarina Fonseca, Dr Hooman Kamel, Dr Patricia Martinez Sanchez, Dr Gerben Plas, Dr
Timolaos Rizos, Dr Wadea Tarhuni, Dr Vincent Thijs, Dr Rolf Wachter, Dr Marianne Zeller,
who provided data from their studies for this analysis.
We would also like to thank Dr Manuel Jesus Gomez-Choco for his help with translations.
None of these individuals had access to the manuscript before submission and they are not
responsible for its contents.
Sources of Funding
St. Mary´s development fund, National Institute for Health Research (NIHR) Imperial
Biomedical Research Centre.
13
Conflicts of Interest/|Disclosures
EK, ADG, SH, UM: None
DG: speaker honoraria, advisory boards: Bayer, Boehringer Ingelheim, BMS, Daiichi-
Sankyo, Pfizer.
PH: research grants BMBF, European Union, Charité, Berlin Chamber of Physicians,
German Parkinson Society, UniversityHospital Würzburg, Robert-Koch-Institute, Charité–
Universitätsmedizin Berlin (MonDAFIS,unrestricted research grant to Charité, Bayer),
University Göttingen (FIND-AFrandomized, unrestricted research grant, University
Göttingen, Boehringer-Ingelheim), UniversityHospital Heidelberg (RASUNOA-prime;
unrestricted research grant , Bayer, BMS, Boehringer, Daiichi Sankyo), outside submitted
work.
RV: Consulting, speaker honoraria, research support: Bayer, Boehringer , Daiichi Sankyo,
BMS, Pfizer, Biogen, Morphosys, Medtronic, Apoplex Medical technologies
14
References
1. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med. 2007;146(12):857-67.2. Kallmunzer B, Breuer L, Hering C, et al. A structured reading algorithm improves telemetric detection of atrial fibrillation after acute ischemic stroke. Stroke. 2012;43(4):994-9.3. Rizos T, Güntner J, Jenetzky E, et al. Continuous stroke unit electrocardiographic monitoring versus 24-hour Holter electrocardiography for detection of paroxysmal atrial fibrillation after stroke. Stroke. 2012;43(10):2689-94.4. Easton JD, Saver JL, Albers GW, et al. Definition and evaluation of transient ischemic attack: a scientific statement for healthcare professionals from the American Heart Association/American Stroke Association Stroke Council; Council on Cardiovascular Surgery and Anesthesia; Council on Cardiovascular Radiology and Intervention; Council on Cardiovascular Nursing; and the Interdisciplinary Council on Peripheral Vascular Disease. The American Academy of Neurology affirms the value of this statement as an educational tool for neurologists. Stroke. 2009;40(6):2276-93.5. Rothwell PM, Giles MF, Chandratheva A, et al. Effect of urgent treatment of transient ischaemic attack and minor stroke on early recurrent stroke (EXPRESS study): a prospective population-based sequential comparison. Lancet. 2007;370(9596):1432-42.6. Lavallee PC, Meseguer E, Abboud H, et al. A transient ischaemic attack clinic with round-the-clock access (SOS-TIA): feasibility and effects. Lancet Neurol. 2007;6(11):953-60.7. Verma A, Cairns JA, Mitchell LB, et al. 2014 focused update of the Canadian Cardiovascular Society Guidelines for the management of atrial fibrillation. Can J Cardiol. 2014;30(10):1114-30.8. Jauch EC, Saver JL, Adams HP, Jr., et al. Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2013;44(3):870-947.9. Guidelines for management of ischaemic stroke and transient ischaemic attack 2008. Cerebrovasc Dis. 2008;25(5):457-507.10. Rizos T, Quilitzsch A, Busse O, et al. Diagnostic work-up for detection of paroxysmal atrial fibrillation after acute ischemic stroke: cross-sectional survey on German stroke units. Stroke. 2015;46(6):1693-5.11. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. J Clin Epidemiol. 2009;62(10):e1-34.12. Higgins JPT AD. Assessing risk of bias in included studies : in Higgins JPT GS, editor: Chichester, UK, Wiley; 2008. 1-53 p.13. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. Bmj. 1997;315(7109):629-34.14. Nyaga VN, Arbyn M, Aerts M. Metaprop: a Stata command to perform meta-analysis of binomial data. Arch Public Health. 2014;72(1):39.15. Barthelemy JC, Feasson-Gerard S, Garnier P, et al. Automatic cardiac event recorders reveal paroxysmal atrial fibrillation after unexplained strokes or transient ischemic attacks. Ann Noninvasive Electrocardiol. 2003;8(3):194-9.16. Callero EC, Martinez-Sanchez P, Gomar DP, et al. Utility a second 24-hours holter monitoring for the diagnosis of paroxismal atrial fibrillation after an acute stroke. Stroke. 2012;43 (2 Meeting Abstracts).17. Christensen LM, Krieger DW, Hojberg S, et al. Paroxysmal atrial fibrillation occurs often in cryptogenic ischaemic stroke. Final results from the SURPRISE study. Eur J Neurol. 2014;21(6):884-9.
15
18. Deelawar S, Manegold JC, Kalyani M, et al. Diagnostic yield of extended-time external loop recording for the detection of paroxysmal atrial fibrillation in patients with cryptogenic stroke. Europace. 2013;15:ii220.19. Dion F, Saudeau D, Bonnaud I, et al. Unexpected low prevalence of atrial fibrillation in cryptogenic ischemic stroke: a prospective study. J Interv Card Electrophysiol. 2010;28(2):101-7.20. Dobbels L, Thijs V. Prospective comparison of continuous cardiac monitoring and Holter on a stroke unit. Eur J Neurol. 2014;21:116.21. Fernandez V, Bejot Y, Zeller M, et al. Silent atrial fibrillation after ischemic stroke or transient ischemic attack: interest of continuous ECG monitoring. Eur Neurol. 2014;71(5-6):313-8.22. Fonseca AC, Brito D, Pinho e Melo T, et al. N-terminal pro-brain natriuretic peptide shows diagnostic accuracy for detecting atrial fibrillation in cryptogenic stroke patients. Int J Stroke. 2014;9(4):419-25.23. Gladstone DJ, Spring M, Dorian P, et al. Atrial fibrillation in patients with cryptogenic stroke. N Engl J Med. 2014;370(26):2467-77.24. Grond M, Jauss M, Hamann G, et al. Improved Detection of Silent Atrial Fibrillation Using 72-Hour Holter ECG in Patients With Ischemic Stroke A Prospective Multicenter Cohort Study. Stroke. 2013;44(12):3357-64.25. Jabaudon D, Sztajzel J, Sievert K, Landis T, Sztajzel R. Usefulness of ambulatory 7-day ECG monitoring for the detection of atrial fibrillation and flutter after acute stroke and transient ischemic attack. Stroke. 2004;35(7):1647-51.26. Kamel H, Navi BB, Elijovich L, et al. Pilot randomized trial of outpatient cardiac monitoring after cryptogenic stroke. Stroke. 2013;44(2):528-30.27. Martinez-Sanchez P, Callero EC, Herranz AC, et al. Detection of paroxysmal atrial fibrillation in patient with acute brain ischemia combining cardiac and holter monitoring: Prevalence and predictors. Stroke. 2012;43 (2 Meeting Abstracts).28. Plas GJ, Bos J, Velthuis BO, Scholten MF, den Hertog HM, Brouwers PJ. Diagnostic yield of external loop recording in patients with acute ischemic stroke or TIA. J Neurol. 2015;262(3):682-8.29. Rizos T, Rasch C, Jenetzky E, et al. Detection of paroxysmal atrial fibrillation in acute stroke patients. Cerebrovasc Dis. 2010;30(4):410-7.30. Rodriguez-Leal C, Toro-Cebada R, Quezada Feijoo M, Cabeza-Letran L. Holter 24-hours ECG and clinical predictors for arrhythmia detection in patients with cardioembolic stroke profile. Thromb Res. 2014;133:S115.31. Sanna T, Diener HC, Passman RS, et al. Cryptogenic stroke and underlying atrial fibrillation. N Engl J Med. 2014;370(26):2478-86.32. Tarhuni W, El-Masri M. A Double-blind randomized controlled trial comparing the effectiveness of external loop monitor and automatic trigger atrial fibrillation memory loop recorder in detecting atrial fibrillation among patients with tia: A pilot study. Crit Pathw Cardiol. 2014;13 (3):130.33. Thakkar S, Bagarhatta R. Detection of paroxysmal atrial fibrillation or flutter in patients with acute ischemic stroke or transient ischemic attack by Holter monitoring. Indian Heart J. 2014;66(2):188-92.34. Ungar A, Rieger G, West T, et al. Atrial fibrillation and treatment changes in cryptogenic stroke patients with an implantable loop recorder for continuous cardiac rhythm monitoring. Eur Heart J. 2013;34:62.35. Ustrell X, Pellise A, Merce J, Vinas J, Mares R. Feasibility of implantable cardiac monitors for the detection of occult atrial fibrillation in embolic cryptogenic stroke. Cerebrovascular Dis. 2012;33:376-7.36. Vivanco Hidalgo RM, Rodriguez Campello A, Ois Santiago A, Cuadrado Godia E, Pont Sunyer C, Roquer J. Cardiac monitoring in stroke units: importance of diagnosing atrial fibrillation in acute ischemic stroke. Rev Esp Cardiol. 2009;62(5):564-7.
16
37. Wachter R, Weber-Kruger M, Seegers J, et al. Age-dependent yield of screening for undetected atrial fibrillation in stroke patients: the Find-AF study. J Neurol. 2013;260(8):2042-5.38. Huedo-Medina TB, Sanchez-Meca J, Marin-Martinez F, Botella J. Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods. 2006;11(2):193-206.39. Hunter JP, Saratzis A, Sutton AJ, Boucher RH, Sayers RD, Bown MJ. In meta-analyses of proportion studies, funnel plots were found to be an inaccurate method of assessing publication bias. J Clin Epidemiol. 2014;67(8):897-903.40. Sterne JA, Egger M. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemiol. 2001;54(10):1046-55.41. Kishore A, Vail A, Majid A, et al. Detection of atrial fibrillation after ischemic stroke or transient ischemic attack: a systematic review and meta-analysis. Stroke. 2014;45(2):520-6.42. Seet RC, Friedman PA, Rabinstein AA. Prolonged rhythm monitoring for the detection of occult paroxysmal atrial fibrillation in ischemic stroke of unknown cause. Circulation. 2011;124(4):477-86.43. Dussault C, Toeg H, Nathan M, Wang ZJ, Roux JF, Secemsky E. Electrocardiographic monitoring for detecting atrial fibrillation after ischemic stroke or transient ischemic attack: systematic review and meta-analysis. Circ Arrhythm Electrophysiol. 2015;8(2):263-9.44. Dutta D, Bowen E, Foy C. Four-year follow-up of transient ischemic attacks, strokes, and mimics: a retrospective transient ischemic attack clinic cohort study. Stroke. 2015;46(5):1227-32.45. Brazzelli M, Chappell FM, Miranda H, et al. Diffusion-weighted imaging and diagnosis of transient ischemic attack. Ann Neurol. 2014;75(1):67-76.46. Sinha AM, Diener HC, Morillo CA, et al. Cryptogenic Stroke and underlying Atrial Fibrillation (CRYSTAL AF): design and rationale. American Heart Journal.160(1):36-41.e1.47. Kernan WN, Ovbiagele B, Black HR, et al. Guidelines for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2014;45(7):2160-236.48. Kimura K, Minematsu K, Wada K, Yonemura K, Nakajima M. Clinical characteristics in transient ischemic attack patients with atrial fibrillation. Intern Med. 2003;42(3):255-8.49. Purroy F, Montaner J, Molina CA, Delgado P, Ribo M, Alvarez-Sabin J. Patterns and predictors of early risk of recurrence after transient ischemic attack with respect to etiologic subtypes. Stroke. 2007;38(12):3225-9.50. Redgrave JN, Coutts SB, Schulz UG, Briley D, Rothwell PM. Systematic review of associations between the presence of acute ischemic lesions on diffusion-weighted imaging and clinical predictors of early stroke risk after transient ischemic attack. Stroke. 2007;38(5):1482-8.51. Gladstone DJ, Dorian P, Spring M, et al. Atrial premature beats predict atrial fibrillation in cryptogenic stroke: results from the EMBRACE trial. Stroke. 2015;46(4):936-41.52. Weber-Kruger M, Groschel K, Mende M, et al. Excessive supraventricular ectopic activity is indicative of paroxysmal atrial fibrillation in patients with cerebral ischemia. PLoS One. 2013;8(6):e67602.53. Scammacca N, Roberts G, Stuebing KK. Meta-Analysis With Complex Research Designs: Dealing With Dependence From Multiple Measures and Multiple Group Comparisons. Rev Educ Res. 2014;84(3):328-64.54. Van den Noortgate W, Lopez-Lopez JA, Marin-Martinez F, Sanchez-Meca J. Three-level meta-analysis of dependent effect sizes. Behav Res Methods. 2013;45(2):576-94.
17
Tables
Table 1. AF detection rates in unselected patients.
*,**same population
CEM: continuous ECG monitoring (inpatient), CEMa: automatically analysed continuous ECG monitoring, MCOT: mobile cardiac output telemetry ELR: external loop recorder,
Study N eligible for analysis
Detection method
Duration of monitoring
Detection rate of new AF, (95% CI)
Dobbels, 2014a* 16 Holter 24h 6.3%
Dobbels, 2014b* 16 CEM 24h 0.0%
Fernandez, 2014 35 CEM ≥24h 5.7%
Grond, 2013 326 Holter 72h 2.8%
Martinez-Sanchez, 2012 114 CEM 48h-72h 0.8%
Rizos, 2012a** 97 Holter 24h 6.2%
Rizos, 2012b** 97 CEM >24h 7.2%
Rizos, 2012c** 97 CEMa >24h 8.2%
Tarhuni, 2014a 80 ELR 14d 3.8%
Tarhuni, 2014b 80 MCOT 14d 3.8%
Thakkar, 2014 10 Holter 24h 0.0%
Wachter, 2013 72 Holter 7d 4.2%
18
Table 2. AF detection rates in selected patients.
Study N Selection criteria MethodOf monitoring
Duration of monitoring
New AF,%
Callero, 2012 19 ECG and 1st 24h Holter: negative
Holter 24h 15.8%
Christensen, 2013 20 Cryptogenic ILR 2y (702d) 15.0%
Dion, 2010 3 Cryptogenic, age<75y, controlled HTN
ILR 14m 0.0%
Fonseca, 2014 5 Cryptogenic Holter 24h (3m and 6m)
0.0%
Gladstone, 2014a 110 Cryptogenic, age>55y
Holter 24h 0.9%
Gladstone, 2014b 98 Cryptogenic, age>55y
ELR 30d 15.3%
Kamel, 2013 8 Cryptogenic, high risk TIA (ABCD2≥4)
MCOT 21d 0.0%
Martinez-Sanchez, 2012b
37 TIA of unknown origin or suspected cardiac embolism
Holter 24h 18.9%
Plas, 2015 15 ECG and 24h telemetry: negative
ELR 24h 13.3%
Rizos, 2010a* 13 Age>60y Holter 24h 0.0%
Rizos, 2010b* 15 Age>60y CEM 48h 26.7%
Sanna, 2014a 21 Cryptogenic, age>40y
ICM 6months 15.0%
Sanna, 2014b 19 Cryptogenic, age>40y
Holter 24 hours 0.0%
*same population
ECG: electrocardiogram, ILR: implantable loop recorder, ELR: external loop recorder, CEM: continuous ECG monitoring, ICM: implantable cardiac monitor, HTN: hypertension
19
Figures
Figure 1. PRISMA flow chart
20
Figure 2. Forrest plot showing proportion of TIA patients diagnosed with AF in all included cohorts
using different modalities of ECG monitoring. ES: effect size, CI: confidence intervals
21
Figure 3. Forrest plot showing proportion of TIA patients diagnosed with AF in selected and
unselected cohorts. ES: effect size, CI: confidence intervals
22
Figure 4. Forrest plot showing proportion of TIA patients diagnosed with AF depending on duration
of ECG monitoring. Duration of monitoring was categorized into <=24 hours, >24 hours to 7 days, >
7 days). ES: effect size, CI: confidence intervals
23
Figure 5. Forrest plot with proportion of TIA patients diagnosed with AF using non-invasive versus
invasive methods of ECG monitoring. ES: effect size, CI: confidence intervals