Risk of Stroke in Patients with Patent Foramen Ovale: An Updated Meta-analysis of Observational Studies Bing Ma, MD,* Guangcong Liu, MD,† Xin Chen, MD,* Jianming Zhang, MD,* Yiting Liu, MD,‡ and Jingpu Shi, MD*
Background: Although patent foramen ovale (PFO) is considered to be associated with cryptogenic stroke (CS), there remains an ongoing disputation on this issue because of unstable results from randomized controlled trials. The aim of this study was to reassess the PFO effect on stroke through observational data. Methods: An electronic search of PubMed, Web of Science, and China National Knowledge Infrastructure (CNKI) were finished. Only case–control studies and cohort studies in Chinese or English were included in the analysis. Then random-effected meta-analysis models were performed to assess the association between PFO and stroke. Results: Twelve case–control studies and 6 cohort studies were eligible. Case–control studies showed strong association between PFO and CS (odds ratio [OR]: 2.94, 95% confidence interval [CI]: 2.06, 4.20; P , .001), but cohort studies failed to demonstrate a significant association (hazard ratio [HR]: 1.28, 95% CI: .91, 1.80; P 5.155). Subgroup analysis revealed that the pooled OR decreased significantly when the region was limited to the United States (OR: 1.52, 95% CI: 1.00, 2.32; P 5 .083). OR of studies that adjusted major confounders was 1.74 (95% CI: 1.22, 2.47; P 5 .119) and highquality studies was 1.68 (95% CI: 1.14, 2.47; P 5 .072). For cohort studies, a weak statistical association was observed in using transesophageal echocardiography (TEE) studies (HR: 1.45, 95% CI: 1.06, 2.01; P 5 .138) and follow-up years less than 4 years’ studies (HR: 1.45, 95% CI: 1.00, 2.09; P 5 .064). Conclusions: Although case–control studies still show a positive effect of PFO on stroke, the results of cohort challenged the credibility. Further trial data are needed to confirm the effect of PFO on stroke. Key Words: Stroke—meta-analysis—PFO—observational studies. Ó 2014 by National Stroke Association
Introduction From the *Department of Clinical Epidemiology, and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, Shenyang; †School of Public Health, China Medical University, Shenyang, Liaoning; and ‡Department of Medical Center, The First Affiliated Hospital, China Medical University, Shenyang, China. Received September 6, 2013; revision received October 11, 2013; accepted October 16, 2013. G.L. and B.M. contributed equally to this study. Address for correspondence to Jingpu Shi, MD, Department of Clinical Epidemiology, Institute of Cardiovascular Diseases and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, No. 155 Nanjing Bei Street, Heping District, Shenyang, Liaoning Province 110001, China. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2014 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2013.10.018
Epidemiologic studies have found patent foramen ovale (PFO) prevalence of 44%-66% of in patients with cryptogenic stroke (CS) as with 27% in autopsy.1 It suggests that in some patients, paradoxical embolism caused by PFO might be the cause of stroke. However, the hypothesized mechanism of paradoxical embolization is only documented in several case reports and remains a presumptive rather than a certain diagnosis.2-4 There have been 2 published meta-analyses that explored the relationship between PFO and CS by pooling only case–control studies.5,6 One of them showed that patients less than 55 years with CS had a PFO prevalence of 6 times greater than that of patients with known cases of stroke.5 The other, which used Bayesian
Journal of Stroke and Cerebrovascular Diseases, Vol. 23, No. 5 (May-June), 2014: pp 1207-1215
1207
B. MA ET AL.
1208
approach, estimated that about two thirds of PFO discovered in patients with CS may be related to the stroke.6 Though the 2 studies both suggested the strong relationship between PFO and CS, the results might be biased because the controls of these studies were patients with known case stroke. Furthermore, recent randomized controlled trials published on The New England Journal of Medicine comparing percutaneous closure with medical therapy failed to find significant difference between the 2 treatment methods.7-9 Although the randomized clinical trials may have been underpowered to detect a benefit of closure, this may have been driven by the original estimated effect size being based on a stronger association than is actually the case. To avoid methodological bias and present the accurate relationship between PFO and stroke, we performed a meta-analysis. We systematically reviewed articles concerning the association between PFO and stroke whose control groups were not the stroke patients. In addition, we also focused on prospective cohort studies.
Materials and Methods In accordance with Meta-analysis of Observational Studies in Epidemiology, we performed a random-effect meta-analysis.10 The study type was not limited to case– control or cohort study. We retrieved articles from both English and Chinese databases (up to August 2013) and also searched by hand. Related articles and potentially relevant articles were also screened.
Search Strategy PubMed and Web of Science were searched using the terms ‘‘stroke,’’ ‘‘PFO OR right-to-left shunt,’’ and ‘‘cohort study OR case-control study’’ from 1988 to August 13, 2013. The China National Knowledge Infrastructure (CNKI) was also searched to involve unpublished data. Study language was restricted to English and Chinese. The reference lists of the each obtained article were manually searched to identify additional potentially relevant citations. Other studies that were involved in previous similar systematic reviews were also included in our study.
Selection Criteria Articles were considered eligible if they met the following criteria: (1) cohort or case–control studies, exploring the association between PFO and stroke, (2) the control group must be derived from nonstroke population, and (3) the PFO and stroke must be clearly diagnosed. Studies whose control group derived from stroke patients were excluded. For cohort studies, a study was excluded if its patients received percutaneous closure. As to dual publications or articles based on the same data set, the most recent publication was included.
Two seasoned authors independently reviewed the articles to judge their eligibility, with all disagreements settled by discussion among all authors till a consensus was made.
Data Extraction and Quality Assessment Two authors extracted data from each included study independently. The following information was extracted: first author’s name, year of publication, country, study design, study population, range for follow-up, PFO diagnostic technique, matched or adjusted variables, adjusted odds ratios (ORs), or hazard ratios (HRs) with 95% confidence interval (CIs). Whenever possible, adjusted ORs or relative risks (RRs) were always preferred. Any disagreements were resolved by discussion till a consensus was reached. To assess the study quality, a 9-star system on the basis of the Newcastle–Ottawa Scale was used.11 In keeping with the Meta-analysis of Observational Studies in Epidemiology guidelines, we did not use the Newcastle– Ottawa Scale to determine study inclusion but to guide the classification of studies by similarities in design. We defined a study with 7 or more awarded stars as highquality study.
Statistical Methods For case–control study, we used 4-fold table data to assess the strength of relationship between PFO and CS. For cohort studies, HR and its 95% CI were used. Heterogeneity of the studies was assessed using the Cochran Q statistic (a P value , .05 was considered statistically significant) and quantified by I2 (I2 . 50% indicates significant heterogeneity; I2 , 25% indicates insignificant heterogeneity).12 Subgroup analyses were also carried out. Sensitivity analysis was followed to confirm our results’ stability. A funnel plot, whose asymmetry strongly indicated the probability of publication bias, was adopted to test publication bias in our study. The tests of Egger and Begg were performed to assess the degree of asymmetry, and we considered a P value less than .05 to be evidence of publication bias.13,14 All statistical analyses were conducted using STATA version 11.0 (Stata, College Station, TX).
Results Literature Search Of the 1559 identified reports, 1501 studies were excluded for failing to meet our inclusion criteria after reviewing their titles and abstract. Fifty-eight studies were full-text reviewed, among which 9 meta-analysis articles were excluded; 13 studies without control group and 18 studies whose control group involved stroke patients were excluded. One study was based on the same data set, and the previous studies were excluded.15 Thus, 18 studies were eligible.15-32 A flow diagram shows the
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
detailed steps of the literature search, which is shown in Figure 1.
Characteristics of the Eligible Studies and Quality Assessment Characteristics of the eligible studies were shown in Table 1. One case–control study containing 2 control group was treated as 2 independent studies.27 A total of 5408 patients (2489 cases/cohort, 3299 controllers) were included in the main analysis. The study design types included prospective cohort studies (n 5 615,23,24,26,28,29), hospital-based case–control studies (n 5 616,18,19,25,27,31), and population-based case–control studies (n 5 717,2022,27,30,32 ). The study population included CS patients (all case-control studies and 2 cohort studies, n 5 15), stroke patients (n 5 2), and healthy population with PFO (n 5 2). Three case–control studies presented result of adjusted OR.19,27,31 Most cohort studies adjusted for a wide of potential confounders and used Cox proportional hazard survival models to assess HR and 95% CI. One cohort study reported relative risks (RRs) instead of HR, which was pooled in the meta-analysis with HR as well. Quality scores of each study were summarized in Table 1. The median score was 6. The median scores of
Figure 1. Reference searched and selection of studies in the meta-analysis.
1209
case–control studies and cohort studies were 6 and 6.5, respectively. High-quality studies that had 7 or more awarded stars included 4 case–control studies22,27,31 and 3 cohort studies.15,24,26
Result of Meta-analysis Random-effect meta-analysis of case–control studies showed a significant positive association between PFO and stroke (OR: 2.94, 95% CI: 2.06, 4.20; P , .001, Fig 2). Data from ‘‘population-based’’ and ‘‘hospital-based’’ case–control studies both showed significant association between PFO and stroke (OR: 2.55, 95% CI: 1.53, 4.25; P , .001 for population based; OR: 3.65, 95% CI: 2.05, 6.50; P , .001 for hospital-based studies, Fig 2). Statistically significant heterogeneity was observed in the study (Q 5 46.81, P , .001, I2 5 74.4%). There was no indication of publication bias from the results of Begg test (P 5 .051) and Egger test (P 5 .88). But the funnel looked asymmetry (Fig 3). So it was difficult to conclude whether the publication bias existed. The result of cohort studies was not consistent with case–control studies. The overall pooled result of cohort studies showed that PFO was not associated with stroke (HR: 1.28, 95% CI: .91, 1.80; P 5.155, Fig 4). No significant heterogeneity was observed in the cohort studies (Q 5 7.88,
1210
Table 1. Characteristics of observation studies included in the meta-analysis
Reference
Region and design
Follow-up years
Diagnostic technique
Webster et al17 Jones et al21 Job et al20 Serena et al22 Petty et al27 Ali Ebrahimi et al30 Pezzini et al32 Lechat et al16 de Belder et al18 Cabanes et al19 Mesa et al25 Petty et al27 Gu et al31
New Zealand, case–control Australia, case–control German, case–control Spain, case–control United States, case–control Iran, case–control Italy, case–control France, case–control England, case–control France, case–control Spain, case–control America, case–control America, case–control
— — — — — — — — — — — — —
TTE TEE TEE TCD TEE TEE 1 TCD TEE 1 TCD TTE TEE TEE TEE TEE TEE
Mas et al23
Europe, cohort
3.2
Homma et al24
United States, cohort
2
Meissner et al26
United States, cohort
5.1
Serena et al28
Spain, cohort
2
Feurer et al29
Germany, cohort
4
Di Tullio et al15
United States, cohort
11
Case/exposure population CS CS CS CS CS CS CS CS CS CS CS CS CS
Control/nonexposure population
Matched or adjusted variables
Healthy population — Healthy population — Healthy population — Healthy population — Healthy population Age, sex, inj, ISRF, ASA Healthy population — Healthy population — Hospital patients — Hospital patients — Hospital patients ASA, mitral valve prolapse Hospital patients — Hospital patients — Hospital patients Age, gender, race, HTN, DM, chol, coronary artery disease, and history of smoking TEE and TTE CS with PFO CS without PFO Age, sex, HTN, DM, chol, smoking TEE Stroke patients Stroke patients DM, HTN, Glasgow with PFO without PFO score, EtOH TEE Normal with PFO Normal without Age, sex, HTN, DM, IHD, PFO smoking, chol, AF, ASA TEE CS with PFO CS without PFO Age, sex, HTN, DM, IHD, smoking, EtOH, migraine, ASA TCD Stroke patients Stroke patients Age, sex, HTN, DM, with PFO without PFO history of MI, AF TTE Normal with PFO Normal without Age, sex, AF, DM, HTN, PFO chol, history of smoking
Adjusted OR/ HR (95% CI)
Total quality
— — — — 1.29 (.78-2.14) — — — — 3.9 (1.5-9.9) — — 1.9 (1.1-3.5)
5 6 3 8 8 3 6 4 3 6 5 8 7
.86 (.31-2.36)
6
.36 (.79-1.95)
7
1.46 (.74-2.88)
7
3.31 (1.48-7.41)
6
.81 (.39-1.69)
6
1.10 (.64-1.91)
8
B. MA ET AL.
Abbreviations: AF, atrial fibrillation; ASA, atrial septal aneurysm; chol, hypercholesterolemia; CS, cryptogenic stroke; DM, diabetes mellitus; EtOH, alcohol intake; HTN, hypertension; ISRF, ischemic stroke risk factors (hypertension, current smoking, atrial fibrillation, ischemic heart disease, congestive heart disease); IHD, ischemic heart disease; inj, number of contrast injections; MI, myocardial infarction; PFO, patent foramen ovale; TCD, transcranial doppler sonography; TEE, transesophageal echocardiography; TTE, transthoracic echocardiography.
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
1211
Figure 2. Meta-analysis of patent foramen ovale and cryptogenic stroke of case–control studies according to population- and hospitalbased studies. Abbreviations: CI, confidence interval; OR, odds ratio. (Color version of figure is available online.)
P 5 .163, I2 5 36.6%). Neither the Begg test (P 5 .851) nor the Egger test (P 5.922) showed that the publication bias existed.
Subgroup and Sensitivity Analyses The effect of PFO on stroke risk in subgroup metaanalyses were shown in Table 2. Compared with overall analysis, there was some difference. In case–control analyses, significant reduction was observed in American population (OR: 1.52, 95% CI: 1.00, 2.32; P 5 .083), studies controlling major confounder (OR: 1.74, 95% CI: 1.22, 2.47; P 5.119), and high-quality studies (OR: 1.68, 95% CI: 1.14, 2.47; P 5 .072). In cohort studies, when stratified by
diagnostic technique, the analysis of transesophageal echocardiography (TEE) group yielded an HR of 1.45 (95% CI: 1.06, 2.01; P 5 .138). The studies whose followup period was less than 4 years also showed an insignificant association between PFO and stroke (HR: 1.45, 95% CI: 1.00, 2.49; P 5 .064). In sensitivity analysis, we recalculated results by excluding each study. The pooled OR of case–control studies ranged from 2.23 (95% CI: 1.85, 2.70) to 3.04 (95% CI: 2.59, 3.59, Fig 5). One study had great influence on the pooled result.32 After excluding this study, the OR dropped to 2.23 but still with statistical difference. In addition, the adjusted summary HR of cohort studies had the lowest HR 1.04 (95% CI: .66, 1.43) and high HR 1.20 (95% CI: .83, 1.57, Fig 6). It indicated that results have good stability.
Discussion
Figure 3.
Funnel plot of case–control study.
This is the first review to summarize the observational evidence on the PFO and stroke discreetly by stratifying analyses on different study design, considering different study design may help us draw our conclusion comprehensively. Our originality lies on the limit of case-control studies that the control population must be derived from population-based or hospital-based cohorts.5 Although the number of included case–control studies was 2-fold larger than the number of cohort studies (12 versus 6), the total number of patients in each group was approximately the same. However, heterogeneity in
1212
B. MA ET AL.
Figure 4. Meta-analysis of patent foramen ovale and stroke in cohort studies. Abbreviations: CI, confidence interval; HR, hazard ratio. (Color version of figure is available online.)
cohort studies was significantly lower than case–control studies and studies of 2 types pointed to different conclusions. Evidence from case–control studies demonstrated a strong association as previous meta-analysis.5 In contrast, the finding was not supported by cohort studies that compared prevalence of stroke between PFO population and non-PFO population. The discrepancy between the findings of the 2 types of studies may be a reflection of observational evidence. As we know, the confounding by selection can create noncomparable population in the case–control and cohort studies, in terms of underlying risk of outcomes of interest for patients. Thus, different group of patients may vary in the distribution of factors that determine both the likelihood that the index event was PFO related and the risk of a recurrent events.33 Further, although all the cohort studies had adjusted for potential confounders, traditional methods of multivariate adjustment were not used in most case–control studies; this also may lead to some bias in statistics. In addition, age as influential factor plays an important role in the PFO-related stroke.21,34 With increasing age, there is an increasing risk for stroke. Kent et al34 found younger patients more susceptible to recurrent stroke. In our review, the average age of case–control studies was significantly lower than that in cohort studies. Besides, several other factors may affect the results, such as differential follow-up years or dropout rates of invasively versus medically treated patients may result in informative censoring. A nonuniform start time can lead to immortal time bias.35 These might lead to the positive results in cohort studies. We had performed subgroup analysis according to the factors that may influence the result.36,37 For case– control studies, the obvious reduction of OR was observed in adjusting major confounder studies and
high-quality studies, which indicated that the association might be overestimated in relative studies with poor quality. Meanwhile, 2 issues that may cause OR reduction should be noted. First and foremost, most high-quality studies had adjusted for factors that might lead to stroke, such as atrial fibrillation, mitral valve prolapse, and hypertension. It was easier to identify PFO effect as an independent factor. Second, these studies had a relatively large sample compared with poor quality studies that provide more powerful statistics. In other subgroup analysis, the strength of the association was very robust and accorded with the precious studies. The weak association was observed in studies using TEE and follow-up year less than 4 years in cohort studies. TEE is considered as the gold standard for the diagnosis of PFO. Its sensitivity and specificity have been reported to be essentially 100%.38 Compared with TEE, the sensitivity of transthoracic echocardiography (TTE) and transcranial doppler sonography (TCD) had been shown to be slightly lower in some study.20,39 This misclassification of exposure status may bring about misclassification bias. Along with the follow-up time, PFO-related stroke may gradually diminish and other factors (hyperlipidemia, hyperglycemia, smoking, etc.) began to take the leading position.8 No association was demonstrated in other group. On the secondary prevention in patients with PFO and stroke, similar discrepancy existed in observational studies and randomized controlled trial (RCT) results.79,40 Kitsios et al40 demonstrated that transcatheter PFO closure were superior to medical therapy for preventing stroke recurrence from observational studies. But the evidence from randomized data did not support either method for the prevention for stroke recurrence.8,9 When the discrepancies were both found in etiology research and preventive treatment studies, it cannot be explained
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
1213
Table 2. Summary risk estimates of association between PFO and stroke Heterogeneity test
Case–control studies Study design Population based Hospital based Region Oceania Europe United States Diagnostic technique TTE TEE TCD Young patients Major confounder control High-quality studies Cohort studies Diagnostic technique TEE non-TEE Outcome Stroke 1 TIA High-quality studies Follow-up year ,4 $4 Study population Healthy people with PFO Stroke patients with PFO
No. of studies
OR/HR (95% CI)
Q
P
I2 (%)
13
2.94 (2.06, 4.20)
46.81
,.001
74.40
7 6
2.55 (1.53, 4.25) 3.65 (2.05, 6.50)
29.23 17.52
,.001 .004
79.50 71.50
2 7 3
3.18 (.73, 13.82) 3.89 (2.79, 5.41) 1.52 (1.00, 2.32)
5.12 8.94 4.99
.024 .177 .083
80.50 32.90 59.90
2 10 1 7 3 4 6
7.99 (3.99, 16.00) 2.55 (1.73, 3.76) 2.94 (2.06, 4.20) 4.05 (3.05, 5.37) 1.74 (1.22, 2.47) 1.68 (1.14, 2.47) 1.27 (.98, 1.64)
.06 36.4
.806 ,.001
0 75.30
6.96 4.25 7.01 7.88
.324 .119 .072 .163
13.90 52.90 57.20 36.60
4 2
1.45 (1.06, 2.01) .99 (.64, 1.53)
5.51 .43
.138 .512
45.60 0
3 3
1.18 (.85, 1.64) 1.23 (.91, 1.68)
.33 .41
.849 .816
0 0
3 3
1.45 (1.00, 2.09) 1.11 (.77, 1.60)
5.51 1.34
.064 .163
63.70 0
2 4
1.23 (.80, 1.88) 1.29 (.93, 1.79)
.4 7.45
.525 .059
0 59.70
Abbreviations: CI, confidence interval; HR, hazard ratio; OR, odds ratio; PFO, patent foramen ovale; TCD, transcranial doppler sonography; TEE, transesophageal echocardiography; TIA, transient ischemic attack; TTE, transthoracic echocardiography.
by coincidence and some specific association must be present. In other words, whether a discovered PFO was caused by stroke depends on their specific conditions.34 There were several limitations in our meta-analysis. First, because not all included studies adjusted for
confounders, the pathogenic effect of PFO on stroke could be affected by other risk factors, such as hypertension, diabetes mellitus, hypercholesterolemia, etc. Second, although we excluded studies that used known case stroke patients as control, the source population of control
Figure 5. Heterogeneity test in case–control studies. Abbreviation: CI, confidence interval. (Color version of figure is available online.)
Figure 6. Heterogeneity test in cohort studies. (Color version of figure is available online.)
1214
group were still different. In cohort studies, the study population was not only derived from primary stroke patients but also from the recurrent stroke patients; therefore, selection bias could not be ignored. Finally, because of limited data, we were unable to analyze the effect of other cofactor clinical characteristics to the PFOrelated stroke. In conclusion, current observational evidence did not yield consistent result about PFO and stroke. It suggested that there were very complex relationship between them and could not highlight the PFO closure as first choice whenever a PFO was found in stroke patients. In the future, more studies are needed to discover the real relationship between PFO and stroke.
References 1. Hagen PT, Scholz DG, Edwards WD. Incidence and size of patent foramen ovale during the first 10 decades of life: an autopsy study of 965 normal hearts. Mayo Clin Proc 1984;59:17-20. 2. Li Y, Margraf J, Kluck B, et al. Thrombolytic therapy for ischemic stroke secondary to paradoxical embolism in pregnancy: a case report and literature review. Neurologist 2012;18:44-48. 3. Chessa M, Drago M, Krantunkov P, et al. Differential diagnosis between patent foramen ovale and pulmonary arteriovenous fistula in two patients with previous cryptogenic stroke caused by presumed paradoxical embolism. J Am Soc Echocardiogr 2002;15:845-846. 4. Rachko M, Safi AM, Yeshou D, et al. Cryptogenic ischemic stroke and paradoxical embolism: should a patent foramen ovale be closed? Case report and literature review. Angiology 2001;52:793-799. 5. Overell JR, Bone I, Lees KR. Interatrial septal abnormalities and stroke: a meta-analysis of case-control studies. Neurology 2000;55:1172-1179. 6. Alsheikh-Ali AA, Thaler DE, Kent DM. Patent foramen ovale in cryptogenic stroke: incidental or pathogenic? Stroke 2009;40:2349-2355. 7. Furlan AJ, Reisman M, Massaro J, et al. Closure or medical therapy for cryptogenic stroke with patent foramen ovale. N Engl J Med 2012;366:991-999. 8. Carroll JD, Saver JL, Thaler DE, et al. Closure of patent foramen ovale versus medical therapy after cryptogenic stroke. N Engl J Med 2013;368:1092-1100. 9. Meier B, Kalesan B, Mattle HP, et al. Percutaneous closure of patent foramen ovale in cryptogenic embolism. N Engl J Med 2013;368:1083-1091. 10. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008-2012. 11. Cota GF, de Sousa MR, Fereguetti TO, et al. Efficacy of anti-leishmania therapy in visceral leishmaniasis among HIV infected patients: a systematic review with indirect comparison. PLoS Negl Trop Dis 2013;7:e2195. 12. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-188. 13. Song F, Gilbody S. Bias in meta-analysis detected by a simple, graphical test. Increase in studies of publication bias coincided with increasing use of meta-analysis. BMJ 1998;316:471.
B. MA ET AL. 14. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-1101. 15. Di Tullio MR, Jin Z, Russo C, et al. Patent foramen ovale, subclinical cerebrovascular disease and ischemic stroke in a population-based cohort. J Am Coll Cardiol 2013. 16. Lechat P, Mas JL, Lascault G, et al. Prevalence of patent foramen ovale in patients with stroke. N Engl J Med 1988;318:1148-1152. 17. Webster MW, Chancellor AM, Smith HJ, et al. Patent foramen ovale in young stroke patients. Lancet 1988; 2:11-12. 18. de Belder MA, Tourikis L, Leech G, et al. Risk of patent foramen ovale for thromboembolic events in all age groups. Am J Cardiol 1992;69:1316-1320. 19. Cabanes L, Mas JL, Cohen A, et al. Atrial septal aneurysm and patent foramen ovale as risk factors for cryptogenic stroke in patients less than 55 years of age. A study using transesophageal echocardiography. Stroke 1993; 24:1865-1873. 20. Job FP, Ringelstein EB, Grafen Y, et al. Comparison of transcranial contrast Doppler sonography and transesophageal contrast echocardiography for the detection of patent foramen ovale in young stroke patients. Am J Cardiol 1994;74:381-384. 21. Jones EF, Calafiore P, Donnan GA, et al. Evidence that patent foramen ovale is not a risk factor for cerebral ischemia in the elderly. Am J Cardiol 1994;74:596-599. 22. Serena J, Segura T, Perez-Ayuso MJ, et al. The need to quantify right-to-left shunt in acute ischemic stroke: a case-control study. Stroke 1998;29:1322-1328. 23. Mas JL, Arquizan C, Lamy C, et al. Recurrent cerebrovascular events associated with patent foramen ovale, atrial septal aneurysm, or both. N Engl J Med 2001; 345:1740-1746. 24. Homma S, Sacco RL, Di Tullio MR, et al. Effect of medical treatment in stroke patients with patent foramen ovale: patent foramen ovale in Cryptogenic Stroke Study. Circulation 2002;105:2625-2631. 25. Mesa D, Franco M, Suarez de Lezo J, et al. Prevalence of patent foramen ovale in young patients with cerebral ischemic accident of unknown origin. Rev Esp Cardiol 2003;56:662-668. 26. Meissner I, Khandheria BK, Heit JA, et al. Patent foramen ovale: innocent or guilty? Evidence from a prospective population-based study. J Am Coll Cardiol 2006; 47:440-445. 27. Petty GW, Khandheria BK, Meissner I, et al. Populationbased study of the relationship between patent foramen ovale and cerebrovascular ischemic events. Mayo Clin Proc 2006;81:602-608. 28. Serena J, Marti-Fabregas J, Santamarina E, et al. Recurrent stroke and massive right-to-left shunt: results from the prospective Spanish multicenter (CODICIA) study. Stroke 2008;39:3131-3136. 29. Feurer R, Sadikovic S, Sepp D, et al. Patent foramen ovale is not associated with an increased risk of stroke recurrence. Eur J Neurol 2010;17:1339-1345. 30. Ali Ebrahimi H, Hamzeaie Moghadam A, Aredestani E. Evaluation of patent foramen ovale in young adults with cryptogenic stroke. ARYA Atheroscler 2011;7:74-77. 31. Gu X, He Y, Li Z, et al. Comparison of frequencies of patent foramen ovale and thoracic aortic atherosclerosis in patients with cryptogenic ischemic stroke undergoing transesophageal echocardiography. Am J Cardiol 2011; 108:1815-1819.
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS 32. Pezzini A, Grassi M, Lodigiani C, et al. Interaction between proatherosclerotic factors and right-to-left shunt on the risk of cryptogenic stroke: the Italian Project on Stroke in Young Adults. Heart 2012;98:485-489. 33. Dahabreh IJ, Kent DM. Index event bias as an explanation for the paradoxes of recurrence risk research. JAMA 2011; 305:822-823. 34. Kent DM, Ruthazer R, Weimar C, et al. An index to identify stroke-related vs incidental patent foramen ovale in cryptogenic stroke. Neurology 2013;81:619-625. 35. Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008;167:492-499. 36. Homma S, Di Tullio MR. Patent foramen ovale and stroke. J Cardiol 2010;56:134-141.
1215
37. Serena J, Jimenez-Nieto M, Silva Y, et al. Patent foramen ovale in cerebral infarction. Curr Cardiol Rev 2010;6:162-174. 38. Schneider B, Zienkiewicz T, Jansen V, et al. Diagnosis of patent foramen ovale by transesophageal echocardiography and correlation with autopsy findings. Am J Cardiol 1996;77:1202-1209. 39. Di Tullio M, Sacco RL, Massaro A, et al. Transcranial Doppler with contrast injection for the detection of patent foramen ovale in stroke patients. Int J Card Imaging 1993; 9:1-5. 40. Kitsios GD, Dahabreh IJ, Abu Dabrh AM, et al. Patent foramen ovale closure and medical treatments for secondary stroke prevention: a systematic review of observational and randomized evidence. Stroke 2012;43:422-431.
Background: Although patent foramen ovale (PFO) is considered to be associated with cryptogenic stroke (CS), there remains an ongoing disputation on this issue because of unstable results from randomized controlled trials. The aim of this study was to reassess the PFO effect on stroke through observational data. Methods: An electronic search of PubMed, Web of Science, and China National Knowledge Infrastructure (CNKI) were finished. Only case–control studies and cohort studies in Chinese or English were included in the analysis. Then random-effected meta-analysis models were performed to assess the association between PFO and stroke. Results: Twelve case–control studies and 6 cohort studies were eligible. Case–control studies showed strong association between PFO and CS (odds ratio [OR]: 2.94, 95% confidence interval [CI]: 2.06, 4.20; P , .001), but cohort studies failed to demonstrate a significant association (hazard ratio [HR]: 1.28, 95% CI: .91, 1.80; P 5.155). Subgroup analysis revealed that the pooled OR decreased significantly when the region was limited to the United States (OR: 1.52, 95% CI: 1.00, 2.32; P 5 .083). OR of studies that adjusted major confounders was 1.74 (95% CI: 1.22, 2.47; P 5 .119) and highquality studies was 1.68 (95% CI: 1.14, 2.47; P 5 .072). For cohort studies, a weak statistical association was observed in using transesophageal echocardiography (TEE) studies (HR: 1.45, 95% CI: 1.06, 2.01; P 5 .138) and follow-up years less than 4 years’ studies (HR: 1.45, 95% CI: 1.00, 2.09; P 5 .064). Conclusions: Although case–control studies still show a positive effect of PFO on stroke, the results of cohort challenged the credibility. Further trial data are needed to confirm the effect of PFO on stroke. Key Words: Stroke—meta-analysis—PFO—observational studies. Ó 2014 by National Stroke Association
Introduction From the *Department of Clinical Epidemiology, and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, Shenyang; †School of Public Health, China Medical University, Shenyang, Liaoning; and ‡Department of Medical Center, The First Affiliated Hospital, China Medical University, Shenyang, China. Received September 6, 2013; revision received October 11, 2013; accepted October 16, 2013. G.L. and B.M. contributed equally to this study. Address for correspondence to Jingpu Shi, MD, Department of Clinical Epidemiology, Institute of Cardiovascular Diseases and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, No. 155 Nanjing Bei Street, Heping District, Shenyang, Liaoning Province 110001, China. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2014 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2013.10.018
Epidemiologic studies have found patent foramen ovale (PFO) prevalence of 44%-66% of in patients with cryptogenic stroke (CS) as with 27% in autopsy.1 It suggests that in some patients, paradoxical embolism caused by PFO might be the cause of stroke. However, the hypothesized mechanism of paradoxical embolization is only documented in several case reports and remains a presumptive rather than a certain diagnosis.2-4 There have been 2 published meta-analyses that explored the relationship between PFO and CS by pooling only case–control studies.5,6 One of them showed that patients less than 55 years with CS had a PFO prevalence of 6 times greater than that of patients with known cases of stroke.5 The other, which used Bayesian
Journal of Stroke and Cerebrovascular Diseases, Vol. 23, No. 5 (May-June), 2014: pp 1207-1215
1207
B. MA ET AL.
1208
approach, estimated that about two thirds of PFO discovered in patients with CS may be related to the stroke.6 Though the 2 studies both suggested the strong relationship between PFO and CS, the results might be biased because the controls of these studies were patients with known case stroke. Furthermore, recent randomized controlled trials published on The New England Journal of Medicine comparing percutaneous closure with medical therapy failed to find significant difference between the 2 treatment methods.7-9 Although the randomized clinical trials may have been underpowered to detect a benefit of closure, this may have been driven by the original estimated effect size being based on a stronger association than is actually the case. To avoid methodological bias and present the accurate relationship between PFO and stroke, we performed a meta-analysis. We systematically reviewed articles concerning the association between PFO and stroke whose control groups were not the stroke patients. In addition, we also focused on prospective cohort studies.
Materials and Methods In accordance with Meta-analysis of Observational Studies in Epidemiology, we performed a random-effect meta-analysis.10 The study type was not limited to case– control or cohort study. We retrieved articles from both English and Chinese databases (up to August 2013) and also searched by hand. Related articles and potentially relevant articles were also screened.
Search Strategy PubMed and Web of Science were searched using the terms ‘‘stroke,’’ ‘‘PFO OR right-to-left shunt,’’ and ‘‘cohort study OR case-control study’’ from 1988 to August 13, 2013. The China National Knowledge Infrastructure (CNKI) was also searched to involve unpublished data. Study language was restricted to English and Chinese. The reference lists of the each obtained article were manually searched to identify additional potentially relevant citations. Other studies that were involved in previous similar systematic reviews were also included in our study.
Selection Criteria Articles were considered eligible if they met the following criteria: (1) cohort or case–control studies, exploring the association between PFO and stroke, (2) the control group must be derived from nonstroke population, and (3) the PFO and stroke must be clearly diagnosed. Studies whose control group derived from stroke patients were excluded. For cohort studies, a study was excluded if its patients received percutaneous closure. As to dual publications or articles based on the same data set, the most recent publication was included.
Two seasoned authors independently reviewed the articles to judge their eligibility, with all disagreements settled by discussion among all authors till a consensus was made.
Data Extraction and Quality Assessment Two authors extracted data from each included study independently. The following information was extracted: first author’s name, year of publication, country, study design, study population, range for follow-up, PFO diagnostic technique, matched or adjusted variables, adjusted odds ratios (ORs), or hazard ratios (HRs) with 95% confidence interval (CIs). Whenever possible, adjusted ORs or relative risks (RRs) were always preferred. Any disagreements were resolved by discussion till a consensus was reached. To assess the study quality, a 9-star system on the basis of the Newcastle–Ottawa Scale was used.11 In keeping with the Meta-analysis of Observational Studies in Epidemiology guidelines, we did not use the Newcastle– Ottawa Scale to determine study inclusion but to guide the classification of studies by similarities in design. We defined a study with 7 or more awarded stars as highquality study.
Statistical Methods For case–control study, we used 4-fold table data to assess the strength of relationship between PFO and CS. For cohort studies, HR and its 95% CI were used. Heterogeneity of the studies was assessed using the Cochran Q statistic (a P value , .05 was considered statistically significant) and quantified by I2 (I2 . 50% indicates significant heterogeneity; I2 , 25% indicates insignificant heterogeneity).12 Subgroup analyses were also carried out. Sensitivity analysis was followed to confirm our results’ stability. A funnel plot, whose asymmetry strongly indicated the probability of publication bias, was adopted to test publication bias in our study. The tests of Egger and Begg were performed to assess the degree of asymmetry, and we considered a P value less than .05 to be evidence of publication bias.13,14 All statistical analyses were conducted using STATA version 11.0 (Stata, College Station, TX).
Results Literature Search Of the 1559 identified reports, 1501 studies were excluded for failing to meet our inclusion criteria after reviewing their titles and abstract. Fifty-eight studies were full-text reviewed, among which 9 meta-analysis articles were excluded; 13 studies without control group and 18 studies whose control group involved stroke patients were excluded. One study was based on the same data set, and the previous studies were excluded.15 Thus, 18 studies were eligible.15-32 A flow diagram shows the
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
detailed steps of the literature search, which is shown in Figure 1.
Characteristics of the Eligible Studies and Quality Assessment Characteristics of the eligible studies were shown in Table 1. One case–control study containing 2 control group was treated as 2 independent studies.27 A total of 5408 patients (2489 cases/cohort, 3299 controllers) were included in the main analysis. The study design types included prospective cohort studies (n 5 615,23,24,26,28,29), hospital-based case–control studies (n 5 616,18,19,25,27,31), and population-based case–control studies (n 5 717,2022,27,30,32 ). The study population included CS patients (all case-control studies and 2 cohort studies, n 5 15), stroke patients (n 5 2), and healthy population with PFO (n 5 2). Three case–control studies presented result of adjusted OR.19,27,31 Most cohort studies adjusted for a wide of potential confounders and used Cox proportional hazard survival models to assess HR and 95% CI. One cohort study reported relative risks (RRs) instead of HR, which was pooled in the meta-analysis with HR as well. Quality scores of each study were summarized in Table 1. The median score was 6. The median scores of
Figure 1. Reference searched and selection of studies in the meta-analysis.
1209
case–control studies and cohort studies were 6 and 6.5, respectively. High-quality studies that had 7 or more awarded stars included 4 case–control studies22,27,31 and 3 cohort studies.15,24,26
Result of Meta-analysis Random-effect meta-analysis of case–control studies showed a significant positive association between PFO and stroke (OR: 2.94, 95% CI: 2.06, 4.20; P , .001, Fig 2). Data from ‘‘population-based’’ and ‘‘hospital-based’’ case–control studies both showed significant association between PFO and stroke (OR: 2.55, 95% CI: 1.53, 4.25; P , .001 for population based; OR: 3.65, 95% CI: 2.05, 6.50; P , .001 for hospital-based studies, Fig 2). Statistically significant heterogeneity was observed in the study (Q 5 46.81, P , .001, I2 5 74.4%). There was no indication of publication bias from the results of Begg test (P 5 .051) and Egger test (P 5 .88). But the funnel looked asymmetry (Fig 3). So it was difficult to conclude whether the publication bias existed. The result of cohort studies was not consistent with case–control studies. The overall pooled result of cohort studies showed that PFO was not associated with stroke (HR: 1.28, 95% CI: .91, 1.80; P 5.155, Fig 4). No significant heterogeneity was observed in the cohort studies (Q 5 7.88,
1210
Table 1. Characteristics of observation studies included in the meta-analysis
Reference
Region and design
Follow-up years
Diagnostic technique
Webster et al17 Jones et al21 Job et al20 Serena et al22 Petty et al27 Ali Ebrahimi et al30 Pezzini et al32 Lechat et al16 de Belder et al18 Cabanes et al19 Mesa et al25 Petty et al27 Gu et al31
New Zealand, case–control Australia, case–control German, case–control Spain, case–control United States, case–control Iran, case–control Italy, case–control France, case–control England, case–control France, case–control Spain, case–control America, case–control America, case–control
— — — — — — — — — — — — —
TTE TEE TEE TCD TEE TEE 1 TCD TEE 1 TCD TTE TEE TEE TEE TEE TEE
Mas et al23
Europe, cohort
3.2
Homma et al24
United States, cohort
2
Meissner et al26
United States, cohort
5.1
Serena et al28
Spain, cohort
2
Feurer et al29
Germany, cohort
4
Di Tullio et al15
United States, cohort
11
Case/exposure population CS CS CS CS CS CS CS CS CS CS CS CS CS
Control/nonexposure population
Matched or adjusted variables
Healthy population — Healthy population — Healthy population — Healthy population — Healthy population Age, sex, inj, ISRF, ASA Healthy population — Healthy population — Hospital patients — Hospital patients — Hospital patients ASA, mitral valve prolapse Hospital patients — Hospital patients — Hospital patients Age, gender, race, HTN, DM, chol, coronary artery disease, and history of smoking TEE and TTE CS with PFO CS without PFO Age, sex, HTN, DM, chol, smoking TEE Stroke patients Stroke patients DM, HTN, Glasgow with PFO without PFO score, EtOH TEE Normal with PFO Normal without Age, sex, HTN, DM, IHD, PFO smoking, chol, AF, ASA TEE CS with PFO CS without PFO Age, sex, HTN, DM, IHD, smoking, EtOH, migraine, ASA TCD Stroke patients Stroke patients Age, sex, HTN, DM, with PFO without PFO history of MI, AF TTE Normal with PFO Normal without Age, sex, AF, DM, HTN, PFO chol, history of smoking
Adjusted OR/ HR (95% CI)
Total quality
— — — — 1.29 (.78-2.14) — — — — 3.9 (1.5-9.9) — — 1.9 (1.1-3.5)
5 6 3 8 8 3 6 4 3 6 5 8 7
.86 (.31-2.36)
6
.36 (.79-1.95)
7
1.46 (.74-2.88)
7
3.31 (1.48-7.41)
6
.81 (.39-1.69)
6
1.10 (.64-1.91)
8
B. MA ET AL.
Abbreviations: AF, atrial fibrillation; ASA, atrial septal aneurysm; chol, hypercholesterolemia; CS, cryptogenic stroke; DM, diabetes mellitus; EtOH, alcohol intake; HTN, hypertension; ISRF, ischemic stroke risk factors (hypertension, current smoking, atrial fibrillation, ischemic heart disease, congestive heart disease); IHD, ischemic heart disease; inj, number of contrast injections; MI, myocardial infarction; PFO, patent foramen ovale; TCD, transcranial doppler sonography; TEE, transesophageal echocardiography; TTE, transthoracic echocardiography.
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
1211
Figure 2. Meta-analysis of patent foramen ovale and cryptogenic stroke of case–control studies according to population- and hospitalbased studies. Abbreviations: CI, confidence interval; OR, odds ratio. (Color version of figure is available online.)
P 5 .163, I2 5 36.6%). Neither the Begg test (P 5 .851) nor the Egger test (P 5.922) showed that the publication bias existed.
Subgroup and Sensitivity Analyses The effect of PFO on stroke risk in subgroup metaanalyses were shown in Table 2. Compared with overall analysis, there was some difference. In case–control analyses, significant reduction was observed in American population (OR: 1.52, 95% CI: 1.00, 2.32; P 5 .083), studies controlling major confounder (OR: 1.74, 95% CI: 1.22, 2.47; P 5.119), and high-quality studies (OR: 1.68, 95% CI: 1.14, 2.47; P 5 .072). In cohort studies, when stratified by
diagnostic technique, the analysis of transesophageal echocardiography (TEE) group yielded an HR of 1.45 (95% CI: 1.06, 2.01; P 5 .138). The studies whose followup period was less than 4 years also showed an insignificant association between PFO and stroke (HR: 1.45, 95% CI: 1.00, 2.49; P 5 .064). In sensitivity analysis, we recalculated results by excluding each study. The pooled OR of case–control studies ranged from 2.23 (95% CI: 1.85, 2.70) to 3.04 (95% CI: 2.59, 3.59, Fig 5). One study had great influence on the pooled result.32 After excluding this study, the OR dropped to 2.23 but still with statistical difference. In addition, the adjusted summary HR of cohort studies had the lowest HR 1.04 (95% CI: .66, 1.43) and high HR 1.20 (95% CI: .83, 1.57, Fig 6). It indicated that results have good stability.
Discussion
Figure 3.
Funnel plot of case–control study.
This is the first review to summarize the observational evidence on the PFO and stroke discreetly by stratifying analyses on different study design, considering different study design may help us draw our conclusion comprehensively. Our originality lies on the limit of case-control studies that the control population must be derived from population-based or hospital-based cohorts.5 Although the number of included case–control studies was 2-fold larger than the number of cohort studies (12 versus 6), the total number of patients in each group was approximately the same. However, heterogeneity in
1212
B. MA ET AL.
Figure 4. Meta-analysis of patent foramen ovale and stroke in cohort studies. Abbreviations: CI, confidence interval; HR, hazard ratio. (Color version of figure is available online.)
cohort studies was significantly lower than case–control studies and studies of 2 types pointed to different conclusions. Evidence from case–control studies demonstrated a strong association as previous meta-analysis.5 In contrast, the finding was not supported by cohort studies that compared prevalence of stroke between PFO population and non-PFO population. The discrepancy between the findings of the 2 types of studies may be a reflection of observational evidence. As we know, the confounding by selection can create noncomparable population in the case–control and cohort studies, in terms of underlying risk of outcomes of interest for patients. Thus, different group of patients may vary in the distribution of factors that determine both the likelihood that the index event was PFO related and the risk of a recurrent events.33 Further, although all the cohort studies had adjusted for potential confounders, traditional methods of multivariate adjustment were not used in most case–control studies; this also may lead to some bias in statistics. In addition, age as influential factor plays an important role in the PFO-related stroke.21,34 With increasing age, there is an increasing risk for stroke. Kent et al34 found younger patients more susceptible to recurrent stroke. In our review, the average age of case–control studies was significantly lower than that in cohort studies. Besides, several other factors may affect the results, such as differential follow-up years or dropout rates of invasively versus medically treated patients may result in informative censoring. A nonuniform start time can lead to immortal time bias.35 These might lead to the positive results in cohort studies. We had performed subgroup analysis according to the factors that may influence the result.36,37 For case– control studies, the obvious reduction of OR was observed in adjusting major confounder studies and
high-quality studies, which indicated that the association might be overestimated in relative studies with poor quality. Meanwhile, 2 issues that may cause OR reduction should be noted. First and foremost, most high-quality studies had adjusted for factors that might lead to stroke, such as atrial fibrillation, mitral valve prolapse, and hypertension. It was easier to identify PFO effect as an independent factor. Second, these studies had a relatively large sample compared with poor quality studies that provide more powerful statistics. In other subgroup analysis, the strength of the association was very robust and accorded with the precious studies. The weak association was observed in studies using TEE and follow-up year less than 4 years in cohort studies. TEE is considered as the gold standard for the diagnosis of PFO. Its sensitivity and specificity have been reported to be essentially 100%.38 Compared with TEE, the sensitivity of transthoracic echocardiography (TTE) and transcranial doppler sonography (TCD) had been shown to be slightly lower in some study.20,39 This misclassification of exposure status may bring about misclassification bias. Along with the follow-up time, PFO-related stroke may gradually diminish and other factors (hyperlipidemia, hyperglycemia, smoking, etc.) began to take the leading position.8 No association was demonstrated in other group. On the secondary prevention in patients with PFO and stroke, similar discrepancy existed in observational studies and randomized controlled trial (RCT) results.79,40 Kitsios et al40 demonstrated that transcatheter PFO closure were superior to medical therapy for preventing stroke recurrence from observational studies. But the evidence from randomized data did not support either method for the prevention for stroke recurrence.8,9 When the discrepancies were both found in etiology research and preventive treatment studies, it cannot be explained
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS
1213
Table 2. Summary risk estimates of association between PFO and stroke Heterogeneity test
Case–control studies Study design Population based Hospital based Region Oceania Europe United States Diagnostic technique TTE TEE TCD Young patients Major confounder control High-quality studies Cohort studies Diagnostic technique TEE non-TEE Outcome Stroke 1 TIA High-quality studies Follow-up year ,4 $4 Study population Healthy people with PFO Stroke patients with PFO
No. of studies
OR/HR (95% CI)
Q
P
I2 (%)
13
2.94 (2.06, 4.20)
46.81
,.001
74.40
7 6
2.55 (1.53, 4.25) 3.65 (2.05, 6.50)
29.23 17.52
,.001 .004
79.50 71.50
2 7 3
3.18 (.73, 13.82) 3.89 (2.79, 5.41) 1.52 (1.00, 2.32)
5.12 8.94 4.99
.024 .177 .083
80.50 32.90 59.90
2 10 1 7 3 4 6
7.99 (3.99, 16.00) 2.55 (1.73, 3.76) 2.94 (2.06, 4.20) 4.05 (3.05, 5.37) 1.74 (1.22, 2.47) 1.68 (1.14, 2.47) 1.27 (.98, 1.64)
.06 36.4
.806 ,.001
0 75.30
6.96 4.25 7.01 7.88
.324 .119 .072 .163
13.90 52.90 57.20 36.60
4 2
1.45 (1.06, 2.01) .99 (.64, 1.53)
5.51 .43
.138 .512
45.60 0
3 3
1.18 (.85, 1.64) 1.23 (.91, 1.68)
.33 .41
.849 .816
0 0
3 3
1.45 (1.00, 2.09) 1.11 (.77, 1.60)
5.51 1.34
.064 .163
63.70 0
2 4
1.23 (.80, 1.88) 1.29 (.93, 1.79)
.4 7.45
.525 .059
0 59.70
Abbreviations: CI, confidence interval; HR, hazard ratio; OR, odds ratio; PFO, patent foramen ovale; TCD, transcranial doppler sonography; TEE, transesophageal echocardiography; TIA, transient ischemic attack; TTE, transthoracic echocardiography.
by coincidence and some specific association must be present. In other words, whether a discovered PFO was caused by stroke depends on their specific conditions.34 There were several limitations in our meta-analysis. First, because not all included studies adjusted for
confounders, the pathogenic effect of PFO on stroke could be affected by other risk factors, such as hypertension, diabetes mellitus, hypercholesterolemia, etc. Second, although we excluded studies that used known case stroke patients as control, the source population of control
Figure 5. Heterogeneity test in case–control studies. Abbreviation: CI, confidence interval. (Color version of figure is available online.)
Figure 6. Heterogeneity test in cohort studies. (Color version of figure is available online.)
1214
group were still different. In cohort studies, the study population was not only derived from primary stroke patients but also from the recurrent stroke patients; therefore, selection bias could not be ignored. Finally, because of limited data, we were unable to analyze the effect of other cofactor clinical characteristics to the PFOrelated stroke. In conclusion, current observational evidence did not yield consistent result about PFO and stroke. It suggested that there were very complex relationship between them and could not highlight the PFO closure as first choice whenever a PFO was found in stroke patients. In the future, more studies are needed to discover the real relationship between PFO and stroke.
References 1. Hagen PT, Scholz DG, Edwards WD. Incidence and size of patent foramen ovale during the first 10 decades of life: an autopsy study of 965 normal hearts. Mayo Clin Proc 1984;59:17-20. 2. Li Y, Margraf J, Kluck B, et al. Thrombolytic therapy for ischemic stroke secondary to paradoxical embolism in pregnancy: a case report and literature review. Neurologist 2012;18:44-48. 3. Chessa M, Drago M, Krantunkov P, et al. Differential diagnosis between patent foramen ovale and pulmonary arteriovenous fistula in two patients with previous cryptogenic stroke caused by presumed paradoxical embolism. J Am Soc Echocardiogr 2002;15:845-846. 4. Rachko M, Safi AM, Yeshou D, et al. Cryptogenic ischemic stroke and paradoxical embolism: should a patent foramen ovale be closed? Case report and literature review. Angiology 2001;52:793-799. 5. Overell JR, Bone I, Lees KR. Interatrial septal abnormalities and stroke: a meta-analysis of case-control studies. Neurology 2000;55:1172-1179. 6. Alsheikh-Ali AA, Thaler DE, Kent DM. Patent foramen ovale in cryptogenic stroke: incidental or pathogenic? Stroke 2009;40:2349-2355. 7. Furlan AJ, Reisman M, Massaro J, et al. Closure or medical therapy for cryptogenic stroke with patent foramen ovale. N Engl J Med 2012;366:991-999. 8. Carroll JD, Saver JL, Thaler DE, et al. Closure of patent foramen ovale versus medical therapy after cryptogenic stroke. N Engl J Med 2013;368:1092-1100. 9. Meier B, Kalesan B, Mattle HP, et al. Percutaneous closure of patent foramen ovale in cryptogenic embolism. N Engl J Med 2013;368:1083-1091. 10. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008-2012. 11. Cota GF, de Sousa MR, Fereguetti TO, et al. Efficacy of anti-leishmania therapy in visceral leishmaniasis among HIV infected patients: a systematic review with indirect comparison. PLoS Negl Trop Dis 2013;7:e2195. 12. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-188. 13. Song F, Gilbody S. Bias in meta-analysis detected by a simple, graphical test. Increase in studies of publication bias coincided with increasing use of meta-analysis. BMJ 1998;316:471.
B. MA ET AL. 14. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-1101. 15. Di Tullio MR, Jin Z, Russo C, et al. Patent foramen ovale, subclinical cerebrovascular disease and ischemic stroke in a population-based cohort. J Am Coll Cardiol 2013. 16. Lechat P, Mas JL, Lascault G, et al. Prevalence of patent foramen ovale in patients with stroke. N Engl J Med 1988;318:1148-1152. 17. Webster MW, Chancellor AM, Smith HJ, et al. Patent foramen ovale in young stroke patients. Lancet 1988; 2:11-12. 18. de Belder MA, Tourikis L, Leech G, et al. Risk of patent foramen ovale for thromboembolic events in all age groups. Am J Cardiol 1992;69:1316-1320. 19. Cabanes L, Mas JL, Cohen A, et al. Atrial septal aneurysm and patent foramen ovale as risk factors for cryptogenic stroke in patients less than 55 years of age. A study using transesophageal echocardiography. Stroke 1993; 24:1865-1873. 20. Job FP, Ringelstein EB, Grafen Y, et al. Comparison of transcranial contrast Doppler sonography and transesophageal contrast echocardiography for the detection of patent foramen ovale in young stroke patients. Am J Cardiol 1994;74:381-384. 21. Jones EF, Calafiore P, Donnan GA, et al. Evidence that patent foramen ovale is not a risk factor for cerebral ischemia in the elderly. Am J Cardiol 1994;74:596-599. 22. Serena J, Segura T, Perez-Ayuso MJ, et al. The need to quantify right-to-left shunt in acute ischemic stroke: a case-control study. Stroke 1998;29:1322-1328. 23. Mas JL, Arquizan C, Lamy C, et al. Recurrent cerebrovascular events associated with patent foramen ovale, atrial septal aneurysm, or both. N Engl J Med 2001; 345:1740-1746. 24. Homma S, Sacco RL, Di Tullio MR, et al. Effect of medical treatment in stroke patients with patent foramen ovale: patent foramen ovale in Cryptogenic Stroke Study. Circulation 2002;105:2625-2631. 25. Mesa D, Franco M, Suarez de Lezo J, et al. Prevalence of patent foramen ovale in young patients with cerebral ischemic accident of unknown origin. Rev Esp Cardiol 2003;56:662-668. 26. Meissner I, Khandheria BK, Heit JA, et al. Patent foramen ovale: innocent or guilty? Evidence from a prospective population-based study. J Am Coll Cardiol 2006; 47:440-445. 27. Petty GW, Khandheria BK, Meissner I, et al. Populationbased study of the relationship between patent foramen ovale and cerebrovascular ischemic events. Mayo Clin Proc 2006;81:602-608. 28. Serena J, Marti-Fabregas J, Santamarina E, et al. Recurrent stroke and massive right-to-left shunt: results from the prospective Spanish multicenter (CODICIA) study. Stroke 2008;39:3131-3136. 29. Feurer R, Sadikovic S, Sepp D, et al. Patent foramen ovale is not associated with an increased risk of stroke recurrence. Eur J Neurol 2010;17:1339-1345. 30. Ali Ebrahimi H, Hamzeaie Moghadam A, Aredestani E. Evaluation of patent foramen ovale in young adults with cryptogenic stroke. ARYA Atheroscler 2011;7:74-77. 31. Gu X, He Y, Li Z, et al. Comparison of frequencies of patent foramen ovale and thoracic aortic atherosclerosis in patients with cryptogenic ischemic stroke undergoing transesophageal echocardiography. Am J Cardiol 2011; 108:1815-1819.
RELATIONSHIP BETWEEN PFO AND STROKE: A META-ANALYSIS 32. Pezzini A, Grassi M, Lodigiani C, et al. Interaction between proatherosclerotic factors and right-to-left shunt on the risk of cryptogenic stroke: the Italian Project on Stroke in Young Adults. Heart 2012;98:485-489. 33. Dahabreh IJ, Kent DM. Index event bias as an explanation for the paradoxes of recurrence risk research. JAMA 2011; 305:822-823. 34. Kent DM, Ruthazer R, Weimar C, et al. An index to identify stroke-related vs incidental patent foramen ovale in cryptogenic stroke. Neurology 2013;81:619-625. 35. Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008;167:492-499. 36. Homma S, Di Tullio MR. Patent foramen ovale and stroke. J Cardiol 2010;56:134-141.
1215
37. Serena J, Jimenez-Nieto M, Silva Y, et al. Patent foramen ovale in cerebral infarction. Curr Cardiol Rev 2010;6:162-174. 38. Schneider B, Zienkiewicz T, Jansen V, et al. Diagnosis of patent foramen ovale by transesophageal echocardiography and correlation with autopsy findings. Am J Cardiol 1996;77:1202-1209. 39. Di Tullio M, Sacco RL, Massaro A, et al. Transcranial Doppler with contrast injection for the detection of patent foramen ovale in stroke patients. Int J Card Imaging 1993; 9:1-5. 40. Kitsios GD, Dahabreh IJ, Abu Dabrh AM, et al. Patent foramen ovale closure and medical treatments for secondary stroke prevention: a systematic review of observational and randomized evidence. Stroke 2012;43:422-431.
