OUP user menu

‘There are no facts, only interpretations’

David R. Holmes, Robert D. Brown
DOI: http://dx.doi.org/10.1093/eurheartj/eht340 3336-3338 First published online: 11 September 2013

This editorial refers to ‘Long-term results of a randomized trial comparing three different devices for percutaneous closure of a patent foramen ovale’, by M. Hornung et al., on page 3362 and ‘Patent foramen ovale transcatheter closure vs. medical therapy on recurrent vascular events: a systematic review and meta-analysis of randomized controlled trials’, by P. Rengifo-Moreno et al., on page 3342

Friedrich Nietzsche's quote in the title could be modified to read ‘There are no facts, only believers’. This modification captures and encompasses many of the issues highlighted in the articles by Hornung et al.1 and Pabb et al.2 related to the complex multifaceted field of patent foramen ovale (PFO) closure and stroke.

There are several accepted observations which could serve as facts:

  1. PFOs are essential in utero and remain frequent during early and adult life, and are documented in ∼20–30% of asymptomatic individuals. The incidence of PFO is significantly increased in patients with cryptogenic stroke.3

  2. Stroke is one of the medical disorders most feared by all patients and ourselves alike.

  3. Among the list of multiple conditions, many of which may occur either individually or together in an individual patient and may be either associated with stroke or can result in a stroke, is the presence of a PFO, particularly a complex PFO.

  4. The causal relationship between PFO and stroke has been controversial.4

  5. A variety of percutaneous devices are available which can close or modify PFOs and abolish right-to-left shunt.

  6. There are a number of meta-analyses of observational studies of PFO closure and outcome.5,6

The recent articles by Hornung et al.1 and Pabb et al.2 shed some light on selected aspects of these observations. As such, they are part of a growing body of information which is complex and often confusing; they add both to the complexity and to the confusion.

Hornung et al.1 report on the longer term results of a randomized trial which compared three different devices: the Amplatzer PFO/ASD occluder (ACA Medical, Plymouth, MN, USA), the Helex septal occluder (W.L. Gore and Associates, Flagstaff, AZ, USA), and the CardioSeal-STARflex device (NMT, Boston, MA, USA), each of which was used in 220 patients. All of these devices had CE mark approval. This was a single-centre study of 660 patients enrolled over a 4 year period during which all procedures were performed by a single operator. One wonders parenthetically about the number of patients during this period of time seen with PFO and cerebral ischaemic symptoms who were treated with conventional anticoagulant or antiplatelet therapy alone or were at least offered that possibility. No patients were lost to follow-up during a mean follow-up time of 59.2 ± 6.0 months. There were several notable findings. Device implantation for PFO closure was technically successful in all interventions. At 5 years of follow-up, complete closure rates ranged from 100% in the Amplatzer group to 96.8% in patients treated with Helex occluders. During follow-up, however, 36% of patients had required a second intervention for closure of a residual shunt. Perhaps most important were significant differences detected in the primary composite endpoint of transient ischaemic attack (TIA), stroke, cerebral death, or paradoxical embolism between the three different devices. This endpoint occurred in 3.0, 4.1, and 5.9%, respectively, for Amplatzer, Helex, and CardioSeal-STARflex (P = 0.042 comparing results with Amplatzer with those with the other two devices). Survival analysis documented that the Amplatzer device was superior in this regard to the Cardio Seal-STARflex (P = 0.01). There was, however, no significant difference between the three devices regarding the combined stroke and death rate or stroke and vascular death. Importantly, the specific endpoint of stroke was not significantly different between the three devices (P = 0.36), but only a total of 12 events occurred at 5 years of follow-up. In terms of TIA, there was a trend towards a reduction with Amplatzer (0.058), but only a total of 10 events were documented. In addition, the frequency of new-onset atrial fibrillation at follow-up varied, being highest in patients in the CardioSeal-STARflex group. This rhythm by itself may result in thrombus in the left atrial appendage (LAA) and the potential for stroke.

The second important observation in the Hornung study is that in this group of patients overall, the frequency of the primary composite endpoint was low (3.8% overall). As the authors note, at the time when the study was initiated, there were insufficient data on event frequency to be able to size the trial appropriately and select an optimal sample size for adequate statistical power analysis. This low event rate continues to affect subsequent trials.2

Both of these findings have important implications—subsequent trials must be planned given the fact that not all devices may be the same in terms of either complete shunt closure or endpoints, which include TIA, stroke, cerebral death, or paradoxical embolus. Secondly, large numbers of patients are needed because the event rates of stroke/TIA are low both in patients treated with closure and in those treated medically.

In the second study by Pabb et al.,2 the authors report a meta-analysis of three randomized clinical trials, all of which have been published.79 Importantly, this meta-analysis lacks individual-level data for analyses; this limitation decreases the significance of potential conclusions. In addition, the medical therapy in all three trials was not standardized, varied tremendously, and the adjunctive therapy in the device group also differed, making a meta-analysis strategy potentially problematic.

The most important issue is that as the authors state ‘all three randomized clinical trials had negative results’.2 Indeed, the final conclusions as reported in each study document that specific finding: (i) ‘Closure of a patent foramen ovale for secondary prevention of cryptogenic embolism did not result in a significant reduction in the risk of recurrent embolic events or death as compared with medical therapy’9; (ii) ‘In patients with cryptogenic stroke or TIA who had a patent foramen ovale, closure with a device did not offer a greater benefit than medical therapy alone for the prevention of recurrent stroke or TIA’;7 and (iii) ‘In the primary intention-to-treat analysis, there was no significant benefit associated with closure of a patent foramen ovale in adults who had had a cryptogenic ischemic stroke. However, closure was superior to medical therapy alone in the prespecified per-protocol and as-treated analyses, with a low rate of associated risks’8 (Table 1). It is a principal of trial design that if the primary trial endpoint is negative, any other analyses should be considered hypothesis-generating. Pabb et al. analyse the data in a variety of different non-primary endpoint strategies including per-protocol analysis which show only borderline statistical significance (P = 0.05) and in subgroup analyses of high risk patients and PFO characteristics which detected a tendency towards benefit in those with significant shunt through the PFO (P = 0.06). Some of the results of these subgroup analyses are not consistent with conventional thinking. For example, in younger patients <45 years there was no significant benefit compared with the older patients when randomized to the closure group. One would have thought that younger patients with presumed fewer co-morbid conditions would have benefited more because the aetiology of a central nervous system ischaemic event would have been more likely to be related to paradoxical embolus via the PFO. This would be consistent with previous case–control and observational studies suggesting a higher occurrence of PFO in cryptogenic stroke, particularly in younger patients. In addition, the presence of an atrial septal aneurysm (defined by some authors as complex anatomy) was not associated with an increased incidence of events in either the device- or medically treated groups, although there was heterogeneity in this regard. Pabb et al.,2 however, reported that the devices were safe although they are associated with an increased incidence of procedural complications and an increased risk of new-onset atrial fibrillation, a risk factor itself for the development of thrombus in the LAA and increased potential for embolic stroke (Table 2).

View this table:
Table 1

Efficacy: intent to treat

EndpointClosure (n = 447)Medical therapy (n = 462)HR (95% CI)P-value
 Composite endpointa23 (5.5)29 (6.8)0.78 (0.45–1.35)0.37
 Stroke12 (2.9)13 (3.1)
 TIA13 (3.1)17 (4.1)0.90 (0.41–1.98)0.79
Carroll80.75 (036–1.55)0.44
 Overallb9/499 (1.8)16/481 (3.3)0.49 (0.22–1.11)0.08
 Overallc7/204 (3.4)11/210 (5.2)0.63 (0.24–1.62)0.34
 Stroke5/165 (3.0)8/163 (4.9)0.58 (0.19–1.76)0.14
  • CI, confidence interval; HR, hazard ratio; TIA, transient ischaemic attack.

  • aComposite of stroke or TIA, death from any cause within 30 days of enrolment, or death from neurological causes between 31 days and 2 years.

  • bComposite of recurrent non-fatal ischaemic stroke, fatal ischaemic stroke, or early death after randomization.

  • cComposite of death, non-fatal stroke, TIA, or peripheral embolism.

View this table:
Table 2

Safety: procedural- or device-related complications

Furlan7 (n = 402)
 Major vascular complications3.2%Haematoma >5 cm, transfusion left atrial perforation, vascular surgical repair, peripheral nerve injury
 Atrial fibrillation5.7%
 Major bleeding2.6%
Carroll8 (n = 499)
 Total serious adverse events4.2%Atrial fibrillation, ischaemic stroke, pericardial effusion/tamponade, residual shunt, bleeding, among others
Meier9 (n = 204)
 Procedural complications1.5%
 Atrial fibrillation2.9%

How might Pabb et al. approach this meta-analysis of three recently published clinical trials, each of which failed to show a significant benefit of transcatheter PFO closure over medical therapy? In the last sentence of the Discussion, the authors suggest that the finding ‘has the potential to affect clinical practice’ and ‘these results might be the best evidence of significant benefit from TC PFO closure’. In the subsequent Conclusions section, the authors approach the conclusions with more delicacy and less certainty, using the phrases ‘TC PFO closure showed a possible benefit in reducing the incidence of recurrent neurological events’ or ‘may be beneficial’. Such delicacy/hesitancy in the conclusions appears entirely appropriate given the negative conclusions of each study. Given the significant limitations of a meta-analysis in general, and the specific limitations of combining clinical trials using different devices and, importantly, very different medical management approaches in both the closure and medical arms, one should be extremely hesitant to utilize such data to change clinical practice. Instead, such data may be used as hypothesis-generating, and the conclusions support the call from other authors for more additional, definitive, adequately powered studies.6

What could be said in conclusion about these two trials?1,2

  1. Not all PFO closure devices are equal; some are more equal than others.

  2. Patients still fear stroke more than perhaps any other medical disorders.

  3. A mechanical fix to a mechanical structural problem makes attractive and intuitive sense to both patients and physicians alike, irrespective of the results of randomized trials.

  4. In this field, trials are very difficult to perform, recruitment is slow, our ability to predict patients at increased risk of events is limited, event rates are low, and investigator and patient biases are very strong.

In the relative absence of hard facts and in the presence of controversial data, we will continue to have varied interpretations and strongly held opinions, but, perhaps more importantly, we will continue to have communities of true believers.

Conflict of interest: none declared.


  • The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.

  • doi:10.1093/eurheartj/eht283.

  • doi:10.1093/eurheartj/eht285.