Why the study?
Does TAVI compared to SAVR have different outcomes when evaluating RCTs versus registry data due to technician bias?
Does TAVI compared to SAVR have different outcomes when evaluating RCTs versus registry data due to technician bias?
Answers are grounded in this paper and cite their sources.
The authors argue that relying solely on RCTs for TAVI vs SAVR guidelines introduces a technician bias, and that real-world registry data demonstrating long-term survival advantages for SAVR should be considered.
“I disapprove of what you say, but I will defend to the death your right to say it” (Voltaire) There is currently a great debate on the value of different types of evidence. Randomized evidence reduces confounding and therefore allows quantification of treatment effects. It is also thought to be generalizable (it has been almost exclusively used to guide treatment recommendations in current valve guidelines1). In contrast, retrospective observational studies suffer from biases that are thought to preclude generalizability. At a recent meeting, a prominent member of the community suggested new ways of dealing with conflict in data interpretation. He drew analogies to ancient warfare, where sometimes individual duels were fought to replace large battlefield encounters. We have applied this idea to the current comparisons of transcatheter aortic valve implantation (TAVI) to surgical aortic valve replacement (SAVR) and suggest a new aspect, which should not be overlooked when interpreting data (ie, a technician bias). In case, readers are of different opinions, we believe in the principles of accurate scientific conduct and would hope that the principle of Voltaire (expressed above) applies. So let’s examine our idea. Bellicose conflicts generally entail direct confrontation of military forces, often inflicting substantial casualties. In ancient times, individual duels were sometimes chosen as replacement for a large battle to avoid mass casualties. Although rare, individual examples prove their existence and partial effectiveness.2,3 However, in case the losing party was not willing to accept the outcome, an all-out battlefield encounter remained an option. Drawing analogies from military battlefields to medical care may sound odd, but slogans such as “fighting heart disease and stroke” elute for instance to “the battle” of health vs disease and sometimes even individual medical disciplines may “fight” each other (eg, if a novel treatment modality competes against established often more invasive treatment options). If the new therapies are not “owned” by the same individuals, issues such as treatment sovereignty, billability and (importantly) the evaluation of treatment efficacy may give rise to conflict. However, in contrast to the military, in medicine, the equivalent to a “battlefield encounter” is no longer possible, once the outcome of a “duel” has found entry into guideline recommendations. If one therapy has “lost,” its treatment potential will no longer be properly assessable, because it will no longer be tested against the “winner.” Thus, for optimal patient care, it is important to make certain that study results; however, they were obtained, actually reflect the true treatment effects. Since no single set of evidence is perfect, it may be wise to consider all available evidence, taking advantages and disadvantages of all types of evidence into account. In the words of our analogy, if duels and large battles come to the same conclusion, the winner is most likely indeed better. If they do not, the inconsistencies may require further investigation. We here suggest that current valve guideline recommendations for the treatment of aortic stenosis are based on outcomes from several individual duels of TAVI vs SAVR (ie, highly selected, randomized controlled trials), at the risk of ignoring outcomes from battlefield encounters of TAVI vs SAVR (ie, all-comers registry analyses). Since outcomes of these RCT (randomized controlled trials)-duels differ significantly from the registry-battles (see Figure 1), a comparison of the 2 treatment options in daily practice may no longer be possible because SAVR will no longer be a recommended option. Illustration of a technician-bias in SAVR-TAVI trials. The left upper panel displays the observed to expected mortality ratios (O/E ratios) calculated from the STS scores and the 30-day mortality provided by the individual primary trial outcomes reports. The Gaussian curve in the right upper panel suggests that an O/E ratio of 1 suggests an average outcome in contemporary practice. The SAVR and TAVI RCT O/E ratios reflect the averages calculated from the values of the left upper panel. Note that both, SAVR and TAVI O/E ratios are above average and that TAVI-values are lower than SAVR-values. The survival curves in the bottom panels show landmark analyses of RCT (left13) and risk-adjusted registry (middle and right14) meta-analyses of TAVI vs SAVR comparisons. Note that an early advantage of TAVI (in RCTs) is not visible in registry evidence (suggesting the technician bias). Note further that long-term outcomes always tilt towards SAVR. Abbreviations: SAVR, surgical aortic valve replacement; TAVI, transcatheter aortic valve implantation. TAVI was introduced into clinical practice as an alternative to SAVR by the systematic conduct of randomized trials starting at prohibitively high to most recently low risk.4 In 2021, TAVI received a primary treatment recommendation for all patients above 75 years of age and all younger patients at high risk (EuroSCORE II > 4%),1 based almost exclusively on randomized evidence. The accumulating new randomized evidence suggests superior short-term outcomes5 and non-inferior longer-term outcomes (mostly 5 years)6 and further expansion of recommendations towards TAVI is expected. Randomized evidence is first choice of guideline committees for its ability to quantify treatment effects and the consideration of being generalizable.1 However, the selectivity of RCTs limit generalizability substantially and Barili et al7 illustrated several RCT biases, which in the context of TAVI-SAVR comparison all appear to favour TAVI. We additionally and newly show here that outcomes from randomized TAVI-SAVR trials appear to be generated by selected, above-average technicians (see Figure 1). Except for the NOTION trials, all surgical RCT-arms show 30-day mortality below predicted mortality from STS-score. Assuming that an O/E ratio of 1.0 reflects average outcome, surgical trial results are therefore superior to the outcomes of every day SAVR practice. Thus, above average technicians seem to have performed the procedure. The values for TAVI were even lower, which may reflect a technical advantage of TAVI since no cardiopulmonary bypass, chest opening and general anaesthesia are required. However, if one inspects reconstructed time-to-event curves from risk-adjusted registry analyses, no short-term difference between TAVI and SAVR in daily practice can be found. In addition, those low-risk trials demonstrating superior short-term outcomes of TAVI, impress by results (eg, on 1-year mortality) that are several fold better than can be found in registry data.8 These findings strongly suggest the presence of a technician bias in RCTs, where better technicians generate superior outcomes and that this bias applies more to TAVI than to SAVR. RCT-results therefore appear to be not only influenced by selecting a special patient-population but also by selecting a special “technician-population.” The impact of what technicians do during a single invasive procedure on long-term outcomes has long been known.9,10 Note that the only randomized trial demonstrating 10-year equality of TAVI and SAVR is the NOTION trial,11 where surgery was the worst of all trials (O/E 1.29, 40% valves sized 19 or 21).12 A landmark meta-analysis of all RCTs up to 2020 suggested a significant survival advantage for SAVR between 3 and 5 years.13 A similar trend can be found in the 6-year outcomes of PARTNER 3 (FDA-Website). If we then consider that among 15 risk-adjusted registry analyses, not a single one favoured TAVI and a meta-analysis suggests an almost 50% mortality risk reduction with SAVR after 5 years or more,14 the impact of such a technician bias may not be trivial. If guideline committees ignore registry data, the above scenario is put into practice, by eliminating the future potential for a “battlefield encounter” of SAVR and TAVI. To finish with our analogy, the “duels” would have (through guideline recommendations) successfully silenced a potentially superior force for the long-term treatment of aortic valve stenosis. We therefore advocate a consistent representation of all evidence types for treatment recommendations. We thank Dr Joseph Woo for providing the spark to this unusual view on currently available evidence. None declared. None declared. No data has been created for this manuscript.
No takes yet. Share an insight, caveat, or question.
Doenst et al. (2025) conducted an editorial in Aortic stenosis. Transcatheter aortic valve implantation (TAVI) vs. Surgical aortic valve replacement (SAVR) was evaluated.
Synapse has enriched 10 related papers. Consider them for comparative context: