Prevalence and Repair of Intraoperatively Diagnosed Patent Foramen Ovale and Association With Perioperative Outcomes and Long-term Survival FREE

The role that patent foramen ovale (PFO) plays in cryptogenic stroke remains controversial. The debate over an association has existed for more than a century, but causal data linking PFO and cryptogenic stroke remain anecdotal.^1- 7 Epidemiological evidence is consistent with an increased risk of stroke associated with PFO but data are not conclusive.^8- 17 The paucity of evidence supporting PFO as the mechanism for cryptogenic stroke has left many questions in the field unanswered, including when PFO repair is appropriate. A number of ongoing randomized trials comparing medical management and percutaneous closure are attempting to answer this question in the cryptogenic stroke population, despite the fact that optimal medical therapy also remains unknown. Incidentally diagnosed PFO in asymptomatic patients presents yet another management challenge that clinicians regularly face.¹⁸

Widespread use of intraoperative transesophageal echocardiography (TEE) during cardiothoracic surgery has made incidental discovery of PFO common.¹⁹ Sukernik and Bennett-Guerrero¹⁸ recommend routine PFO closure when almost no alteration of the surgical plan is required, such as during mitral or tricuspid valve surgeries. It has also been suggested that PFO be repaired when the probability of postoperative right-to-left shunting is significant.^20- 21 Clearly the increased morbidity and mortality associated with intraoperative repair must be weighed against the risks of cyanosis or postoperative stroke.

A recent survey of cardiothoracic surgeons in the United States with a 64% response rate demonstrated a high degree of variability in management of intraoperatively discovered PFO.²² For example, during planned on-pump bypass surgery, 27.9% of responders stated they always closed intraoperatively discovered PFO while 10.3% never did. Interestingly, 11% always converted a planned off-pump procedure to on-pump to close the defect, and this rate rose to 96% if the patient had a history of possible paradoxical embolism. In light of these data, we sought to examine the prevalence of intraoperatively diagnosed PFO in cardiothoracic surgery patients and investigate the relationship of repair on perioperative outcomes and long-term survival.

We reviewed the intraoperative transesophageal echocardiograms of patients undergoing cardiothoracic surgery between 1995 and 2006 at the Cleveland Clinic, Cleveland, Ohio. Data were matched with operative records from the Cleveland Clinic Cardiovascular Information Registry to obtain standardly collected demographic features, surgical management characteristics, and perioperative outcomes. Race/ethnicity was self-identified as either Asian, black, Hispanic, Native American, other, unknown, or white at the time of surgery. Long-term survival was assessed using the Social Security Death Index. The study was conducted with institutional review board approval.

During the study period, 41 578 cardiothoracic surgeries were performed and 14 165 entries (34%) had available intraoperative TEE data both preoperatively and postoperatively. A total of 1046 patients were excluded because of previous diagnosis of PFO or atrial septal defect (ASD). Patients with ASD noted postoperatively but not preoperatively (n = 27) were also excluded from analysis, because it was assumed these defects resulted from septal manipulation during surgery. A single reviewer (R.A.K.) reviewed a random sampling of 100 study echocardiograms for quality control and agreed with the reported PFO interpretation in each study. Medical records and operative notes were reviewed for entries missing PFO status or intraoperative repair status to recover missing data. The final data set comprised 13 092 surgeries (Figure 1). The frequency of the different types of surgeries performed are listed in Table 1.

Postoperative stroke and all-cause hospital death were the primary outcome measures of the study. Length of stay from surgery to discharge, length of stay in intensive care units (ICUs), and total time spent on cardiopulmonary bypass were included as secondary outcome measures. Surgical complications that could result from prolonged operations, such as postoperative myocardial infarction, bleeding, renal failure, septicemia, and circulatory arrest, were also included. Established risk factors for stroke, such as previous stroke or transient ischemic attack (TIA), atrial septal aneurysm, aortic arch atheroma, atrial fibrillation, and left atrial dilation, were included along with other comorbidities as potential confounders in propensity analyses.^14,23- 33

Two perioperative analyses were performed. The first analysis stratified the cohort by presence of intraoperatively diagnosed PFO, and the second analysis stratified patients with intraoperatively diagnosed PFO by whether patients underwent PFO repair. Group comparisons were made using χ² tests for categorical variables and Wilcoxon rank sum tests for continuous variables. Multivariable logistic regression was used to evaluate comparator groups. Bootstrap bagging methods were used to determine reliable predictors of PFO in the first analysis and PFO repair in the second analysis with random resampling and automated stepwise selection (with entry criteria of P ≤ .07 and retention criteria of P ≤ .05).^34- 35 Variables or clusters of variables that entered more than 50% of 1000 models were chosen for the final model.

Once a set of reliable predictors was found, the models were augmented with the most reliable predictors from clinically relevant groups including established risk factors to form a saturated model. Variables used in propensity models included demographic characteristics, risk factors for stroke, preoperative health status, valve pathology, comorbidities, and surgery details. A propensity score was calculated for each patient by solving the saturated model for the probability of PFO detection in the first comparison and the probability of repair in the second comparison.³⁶ The propensity scores from the saturated models were used to perform greedy matching.³⁷ Long-term follow-up analyses were performed in a similar manner and represented in Kaplan-Meier survival curves compared using log-rank tests.³⁸ By using propensity analyses and matching methods, we were able to eliminate bias while still maintaining adequate power to detect differences in complication rates. Data management and statistical analysis were performed in SAS version 9.1 (SAS Institute Inc, Cary, North Carolina).

Patent foramen ovale was intraoperatively discovered in 2277 patients (17%). Table 2 demonstrates a nearly constant rate of intraoperative PFO diagnosis during the study period with a 95% confidence interval of 17% to 18%. However, the rate of repair demonstrates an increasing trend, reaching a peak of almost 40% in 2003 (Table 2).

Risk factors for stroke, including history of stroke or TIA (11% vs 11%, P = .85), aortic arch atheroma (71% vs 71%, P = .89), hypertension (65% vs 67%, P = .32), atrial fibrillation (10% vs 11%, P = .22), and smoking status (58% vs 58%, P = .60), were similar between those with and without PFO (Table 3). Intraoperative PFO was more likely to occur in patients who were older (mean [SD] age, 63.5 [13] vs 62.9 [13] years; P = .03), were white (91% vs 88%, P < .001), and had atrial septal aneurysm (5% vs 1%, P < .001). Patent foramen ovale was also more likely to be discovered in more recent surgeries (mean [SD] time from beginning of study to surgery, 6.9 [3] vs 6.4 [3] years; P < .001), in tricuspid valve repairs (9% vs 7%, P = .002), and with 2 of the 13 surgeons included in the study.

Of patients with incidentally discovered PFO, 639 (28%) underwent surgical repair, nearly all of which were suture closures (97%). Patients undergoing repair were more likely to be women (42% vs 33%, P < .001), be younger (mean [SD] age, 61.1 [14] vs 64.4 [13] years; P < .001), be undergoing mitral or tricuspid valve surgery (51% vs 32%, P < .001), have a history of stroke or TIA (16% vs 10%, P < .001), have a dilated left atrium (61% vs 51%, P < .001), and have atrial fibrillation (13% vs 10%, P = .03) (Table 4). Three surgeons demonstrated increased rates of PFO closure compared with the other 10. Patients who had PFO repaired also had fewer comorbidities, including hypertension, previous myocardial infarction, smoking, peripheral vascular disease, and carotid artery disease.

The perioperative outcomes between those with incidentally discovered PFO and those with no PFO are shown in Table 5. After we performed propensity matching to control for differences between comparator groups, patients with intraoperatively diagnosed PFO had similar rates of in-hospital stroke (2.3% vs 2.3%, P = .84) and hospital death (3.4% vs 2.6%, P = .11). Length of hospital stay (mean [SD] time, 12.7 [14.0] vs 12.1 [11.7] days; P = .21) and days spent in the ICU (mean [SD] time, 3.5 [7.7] vs 3.1 [6.2] days; P = .70) were also similar between those with intraoperatively diagnosed PFO and those without. Secondary outcomes, including myocardial infarction, bleeding, renal failure, septicemia, and cardiac arrest, also demonstrated no differences between groups. However, patients with PFO were exposed to cardiopulmonary bypass longer than those without intraoperatively diagnosed PFO (mean [SD] time, 110 [46.4] vs 104 [41] min; P = .001).

Perioperative outcomes for patients undergoing intraoperative repair compared with those who did not undergo repair are shown in Table 6. The only difference noted between the 2 groups was the rate of in-hospital stroke, which was 2.8% in the repaired group vs 1.2% in the unrepaired group (P = .04), representing 2.47-times greater odds of having in-hospital stroke (95% confidence interval, 1.02 to 6.00). The rate of hospital deaths (2.5% vs 3.2%, P = .50), hospital length of stay (mean [SD] time, 12.2 [11.8] vs 12.3 [15.5] days; P = .88), ICU length of stay (mean [SD] time, 3.4 [7.4] vs 2.8 [7.6] days; P = .26), and time on cardiopulmonary bypass (mean [SD] time, 107 [45] vs 104 [45.6] days; P = .08) were all similar.

There were 66 228 patient years available for long-term survival analysis. Mean [SD] time for follow-up among survivors was 5.6 [3.1] years (median, 5.7 years). Nine hundred fifty-six patients (7.3%) were followed up for more than 10 years (Figure 2). Before matching, survival rates in patients without PFO at 2, 4, 6, 8, and 10 years were 90%, 85%, 78%, 70%, and 63%, respectively, and in patients in the PFO group, 89%, 83%, 76%, 68%, and 60%, respectively. There was no significant difference in the 2 groups (log-rank P = .06). After matching, survival at 2, 4, 6, 8, and 10 years was 89%, 84%, 77%, 70%, and 65%, respectively, in patients without PFO and 89%, 83%, 76%, 68%, and 60%, respectively, in the PFO group, still with no significant difference in the 2 groups (log-rank P = .40).

Before matching, survival in patients with PFO and no operative repair at 2, 4, 6, 8, and 10 years was 89%, 82%, 75%, 66%, and 59%, respectively, and in patients undergoing repair of PFO, 87%, 85%, 80%, 74%, and 66%, respectively. These differences were statistically significant (log-rank P = .03). However, after matching, survival at 2, 4, 6, 8, and 10 years was 91%, 83%, 74%, 65%, and 63%, respectively, in patients with PFO and no operative repair and 89%, 85%, 80%, 73%, and 67%, respectively, in the repair group, with the differences being nonsignificant (log-rank P = .12).

Our population of patients undergoing cardiothoracic surgery was markedly different from those of prior studies investigating the relationship between PFO and stroke. Many patients in our study had comorbid conditions and risk factors for stroke, while past studies have chosen randomly selected healthy subjects or subjects with a history of ischemic stroke.^14- 16 Our population, however, was uniquely suited to examine the perioperative impact of intraoperatively diagnosed PFO and its repair.

We observed the incidence of intraoperative PFO to be 17%. Since our analysis specifically excluded patients with known septal defects, our reported rate of intraoperative PFO underestimates the true prevalence in our initially selected population. If we include the 1046 patients excluded for a previously established diagnosis of either PFO or ASD (with the former expected to outnumber the latter by more than 500:1), the prevalence of PFO in this population becomes 23.5%, which is in line with prior autopsy analysis that estimated the prevalence of PFO in the general population to be as high as 27%.³⁹ Of note, the prospective study conducted by Meissner and colleagues¹⁵ found the prevalence to be 24% while the study by Di Tullio and colleagues¹⁶ found the prevalence to be 15%. Our analysis cannot be directly compared with the latter study, however, because of our use of TEE, which is now considered to be the gold standard in PFO diagnosis.

A consistent prevalence of intraoperatively diagnosed PFO throughout the study suggests that we did not capture a highly selected cohort for identifying PFO. Atrial septal aneurysm was present in approximately 5% of patients with PFO and 1% of patients without PFO. This is consistent with the study from Meissner and colleagues,¹⁵ who reported septal aneurysm in 4% of patients with PFO and 1% of patients without. Aortic arch atheroma was present in more than 70% of the study population, and severe aortic arch atheroma was present in 5%, appropriate frequencies given our aged patient population.

Our data indicate that the rate of intraoperative PFO diagnosis remained fairly constant during the study period, but the rate of PFO repair increased, peaking in the early 2000s. We attribute this trend to increased attention and awareness of PFO as a potential mechanism for stroke, similar to the recent proliferation of percutaneous PFO closure procedures.⁴⁰

Our data show that surgeons were more likely to repair PFO in patients who were younger, were undergoing mitral or tricuspid surgery, had left atrial dilation, or had a history of prior stroke or TIA. However, we also found that patients with incidental PFO were no more likely to have previously experienced a stroke or TIA than patients without PFO. Repair tended to occur in patients with fewer comorbidities, including hypertension, smoking, myocardial infarction, peripheral vascular disease, and carotid disease. These observations are consistent with current opinions regarding the most appropriate patients in whom to repair PFO.¹⁸ The decision of whether to close an incidental PFO is difficult since the optimal management is largely unknown.⁴¹ The surgeon must balance the additional risks from changes in the surgical plan necessary to close a defect with the potential long-term complications, such as paradoxical embolic stroke, if left intact.

Mean length of stay from surgery to discharge was approximately 12 days across the study population. Presence of incidental PFO was only linked to increased time on cardiopulmonary bypass but was not associated with poorer outcomes. The risks of hospital death, days spent in the ICU, and time spent on cardiopulmonary bypass were similar for patients with PFO and those without PFO. Repair of PFO did not appear to change the risk of in-hospital death but was associated with an increased risk of stroke after controlling for differences between comparator groups. Long-term survival analysis also demonstrated no difference between those with intraoperatively diagnosed PFO and those without. Furthermore, long-term mortality appeared similar between those undergoing repair and those whose PFO was left unrepaired.

One might argue that no long-term difference was detected because surgeons were able to properly select patients undergoing repair, but this seems improbable given our extensive propensity-matched analysis. In contrast, we feel these data suggest that asymptomatic PFO in our population was likely a benign entity and repair might have increased the risk of postoperative stroke.

It was standard protocol to obtain intraoperative TEE for the cardiothoracic surgical procedures reviewed. We cannot ensure, however, that full studies were performed in every case (although our sampling suggests that they were), nor can we be assured that PFO detection was rigorously pursued in every case (ie, by using bubble contrast studies or provocative Valsalva maneuvers). Furthermore, we are unaware of what medications our patients were taking before or after surgery. The effect of stopping or starting anticoagulation or antiplatelet therapy could therefore not be assessed and might also serve as a confounder.

In summary, PFO is commonly detected during intraoperative imaging at the time of cardiothoracic surgery. When incidentally discovered, it appears to have a benign short-term and long-term clinical course. While the number of events is small, there was no clear benefit of closure on short-term perioperative outcomes or longer-term mortality. The finding that repair may increase postoperative stroke risk should discourage routine surgical closure and foster further investigation to delineate whether there is any benefit in terms of long-term stroke prevention and which patients might benefit from this intervention.