Programmed Ventricular Stimulation for Risk Stratification in Patients with Tetralogy of Fallot: A Bayesian Perspective

Paul Khairy


Nat Clin Pract Cardiovasc Med. 2007;4(6):292-293. 

Tetralogy of Fallot (TOF) accounts for 10% of congenital heart defects and is the most common form of cyanotic heart disease beyond infancy.[1] Although patients with repaired TOF have an excellent prognosis in the current era, ventricular arrhythmias can arise decades after corrective surgery, and sudden cardiac death (SCD) is the most common cause of mortality late after repair.[2,3,4] Several studies have identified noninvasive predictors of SCD in patients with TOF, which have contributed substantially to clinical risk assessment.[2,5,6,7] Noninvasive factors alone, however, have thus far lacked sufficient predictive accuracy to reliably identify high-risk subgroups and guide management decisions.

Electrophysiological testing with programmed ventricular stimulation (PVS) has previously been proposed as a method of risk stratification in patients with TOF. In a multicenter study, my group demonstrated that inducible ventricular tachycardia is associated with clinical ventricular arrhythmias and SCD in patients with TOF, independent of known noninvasive risk factors.[8] Nevertheless, the exact role of PVS remains contentious. Opinions diverge as to which subgroups of patients with TOF merit electrophysiological assessment and whether this should become a routine screening method. The current literature does not provide definitive responses to such controversies, as prospective, electrophysiology-guided, implantable cardioverter-defibrillator (ICD) trials have yet to be conducted in patients with TOF. Given these unresolved issues, caregivers grapple with decisions regarding risk assessment and SCD prevention.

I propose that a Bayesian perspective can provide some guidance on this matter. Bayesian analysis offers a logical, quantitative framework for probabilistic reasoning. In the Bayesian paradigm, current knowledge about pretest estimates is combined with results of diagnostic tests to generate post-test probabilities for individual patients. Bayes' theorem dictates that pretest odds [pretest probability ÷ (1 - pretest probability)] multiplied by positive likelihood ratios [sensitivity ÷ (1 - specificity)] yields post-test odds, which can be transformed into post-test probabilities [post-test odds ÷ (1 + post-test odds)]. Pretest odds of no disease divided by negative likelihood ratios [specificity ÷ (1 - sensitivity)] generate post-test odds of no disease. Relevant to this analysis, my group has previously quantified the following test characteristics for PVS in TOF—sensitivity 77.4%, specificity 79.5%, positive likelihood ratio 3.77, and negative likelihood ratio 0.28.[8]

More than 10 randomized studies have now assessed effectiveness of ICDs in patients without clinical sustained ventricular arrhythmias, but thought to be at high-risk for such life-threatening events (i.e. primary prevention studies). Four of these studies found no survival benefit with ICDs. Of the studies that did report a mortality reduction, the population enrolled in the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT)[9] was at lowest risk, with a ∼3.5% (3.44%) annual incidence of SCD among control subjects. This estimate is derived from the reported actuarial 5-year mortality of 36.1%, which corresponds to an all-cause mortality rate of 8.6% per year, (see supplementary information [1] online) with SCD accounting for 40% of all deaths (40% of 8.6% = 3.44%).[9] In other words, benefit from ICDs has yet to be demonstrated if the baseline population risk for SCD is less than 3.5% per year.

A population-based study of patients with TOF revealed a 0.15% annual incidence of SCD.[10] Applying Bayes' theorem, if all patients with TOF were routinely screened, a positive study would identify a subgroup with a 0.56% annual risk of SCD (see supplementary information [2] online). As this estimate remains markedly inferior to our previously determined 3.5% threshold, ICD implantation would not be advisable in all patients with TOF in whom ventricular tachycardia can be induced. Consequently, routine screening with PVS is not recommended.

To identify patients with TOF in whom inducing ventricular tachycardia would increase the risk category sufficiently to contemplate ICD implantation, the minimum pretest probability for SCD should be 0.95%, which corresponds to a positive post-test probability of 3.5%. As this annual SCD incidence is over sixfold greater than in the population of patients with TOF at large, noninvasive predictors could be ideally suited for initial risk stratification. For example, a QRS duration longer than 180 ms increases the baseline risk for SCD 2.3-fold.[2] Alone, this risk factor is insufficient to justify further screening with PVS. A combination of static and dynamic noninvasive risk factors, such as age, transannular patch, nonsustained ventricular tachycardia, syncope, and hemodynamic or echocardiographic data is, however, likely to identify a subgroup with a sufficiently high pretest probability to warrant further stratification with electrophysiological testing. The Pediatric and Congenital Electrophysiology Society is currently completing a multicenter study with the objective of establishing such a noninvasive risk score. At the other end of the spectrum, if the post-test probability of SCD remains greater than 3.5% per year despite noninducible ventricular tachycardia, an ICD should be considered on the basis of noninvasive risk factors alone. This situation would arise in patients with TOF if the pretest probability of SCD exceeded 11.5% per year (see supplementary information [3] online).

This Bayesian approach is fraught with assumptions. It draws upon the larger body of knowledge regarding ICDs in other settings. It also presumes that patients with TOF at similar risk of SCD to patients enrolled in primary prevention studies will derive equivalent benefit from ICDs. A multitude of factors inherent to this supposition could differ amongst the two populations of patients, including other causes of mortality. Interestingly, however, both SCD-HeFT[9] and the population-based TOF study[10] noted a similar proportion of nonsudden deaths. Furthermore, of the 16.8% of patients with a positive combined end point of clinical ventricular tachycardia or SCD in the TOF PVS study,[8] nearly 50% had clinical sustained ventricular tachycardia not associated with SCD. Equal weights were attributed to all events in generating test characteristics for PVS. Finally, this analysis supposes that the role of PVS is limited to risk stratification for ICD implantation and does not consider other potential indications such as diagnostic or transcatheter ablation purposes. Importantly, this exercise was undertaken to provide some framework for decision-making based on current knowledge, and is not a substitute for more rigorous formal investigations.

In summary, if the objective is to assess appropriateness for ICD implantation, this Bayesian perspective indicates that PVS need not be routinely performed to risk-stratify all patients with TOF. Three pretest categories based on annual risk of SCD should be defined by noninvasive determinants: low (≤ 1.0%); intermediate (1.0–11.5%); and high (≥11.5%). These categories can then be used for ultimate binary post-test categorization (i.e. <3.5%, ≥3.5%). Patients in low and high-risk categories would not derive benefit from PVS, but those at intermediate risk would. Conceptually, patients at intermediate risk could be subdivided further into those who would shift to 'high risk' following a positive test (1.0–3.5% annual pretest probability of SCD) or to 'low risk' in light of a negative test (3.5–11.5% annual pretest probability of SCD).

CLICK HERE for subscription information about this journal.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: