Fetal aortic valvuloplasty: investigating institutional bias in surgical decision-making

Fetal aortic valvuloplasty may prevent the progression of aortic stenosis to hypoplastic left heart syndrome and allow biventricular rather than univentricular postnatal treatment. This study aimed to investigate whether blinded simulation of a multidisciplinary team approach aids interpretation of multicenter data to uncover institutional bias in postnatal decision-making following fetal cardiac intervention for aortic stenosis.

The study included 109 cases of prenatally diagnosed aortic stenosis from 13 European countries, of which 32 had undergone fetal cardiac intervention. The multidisciplinary team, blinded to fetal cardiac intervention, institutional location and postnatal treatment, retrospectively assigned a surgical pathway (biventricular or univentricular) based on a review of recorded postnatal imaging and clinical characteristics. The team's decisions were the numerical consensus of silent voting, with case review when a decision was split. Funnel plots showing concordance between the multidisciplinary team and the local team's surgical choice (first pathway) and with outcome (final pathway) were created.

In 105 cases the multidisciplinary team reached a consensus decision regarding the surgical pathway, with no decision in four cases because the available imaging records were inadequate. Blinded multidisciplinary team consensus for the first pathway matched the decision of the surgical center in 93/105 (89%) cases, with no difference in agreement between those that had undergone successful fetal cardiac intervention (n = 32) and no (n = 74) or unsuccessful (n = 3) valvuloplasty (no fetal cardiac intervention) (κ = 0.73 (95% CI, 0.38-1.00) vs 0.74 (95% CI, 0.51-0.96)). However, funnel plots comparing multidisciplinary team individual decisions with those of the local teams displayed more discordance (meaning biventricular-univentricular conversion) for the final surgical pathway following fetal cardiac intervention than they did for cases without such intervention (36/74 vs 34/130; P = 0.002), and identified one outlying center.

The use of a blinded multidisciplinary team to simulate decision-making and presentation of data in funnel plots may assist in the interpretation of data submitted to multicenter studies and permit the identification of outliers for further investigation. In the case of aortic stenosis, a high level of agreement was observed between the multidisciplinary team and the surgical centers, but one outlying center was identified.