Reporting surgical outcomes of left-sided endocarditis: can patient features and analysis methods matter more than prosthesis type?

Evidence emerging from longitudinal clinical studies can importantly support our daily surgical practice, guiding decision-making processes. When left-sided valve infective endocarditis is concerned, one of the critical aspects is the choice of the valve substitute. With this respect, to be really helpful, longitudinal studies should focus specifically on the outcomes most relevant to prosthetic choice and apply appropriate balancing statistical strategies to adjust for potential inter-group differences. Although having the merit of reviving the dilemma of whether a prosthetic type (mechanical or biological) is advantageous over the other in the specific setting of left-sided endocarditis, the study by Kahrovic et al. [1] eventually failed to give a convincing answer and concluded that prospective randomized controlled trials are needed. Compared to other valve diseases requiring surgery, the clinical profile of the patients presenting with infective endocarditis is unique, including local features predisposing to valve infection, preoperative acute complications, systemic conditions or general characteristics (not only clinical but also social/behavioral) that make them prone to relapse/recurrence. In our opinion, the most relevant and aetiology-specific long-term outcomes to focus on in this setting should be the risk of recurrence [prosthetic valve endocarditis (PVE)] and the rate of reoperation for recurrence. Indeed, PVE is the most dreaded complication in the follow-up of surgically treated endocarditis, characterized by poorer natural and postoperative prognosis [2]. In the retrospective study by Kahrovic et al. [1], in the subgroup of 88 patients younger than 55 years, the ones receiving a mechanical prosthesis experienced a better overall survival compared to recipients of a bioprosthesis, and a lower risk of a composite event (including death, cerebrovascular events and valve-related reoperation), whereas over this age cut-off the two substitute types seemed to perform similarly. However, these results appear quite intuitive and might reflect the performance of the 2 valve substitutes in the general population, regardless of aetiology, rather than in the specific setting of left-sided endocarditis. Fifty-five years is quite low a cut-off: it cannot be excluded that subdividing the study population differently (e.g. 65 or 70 years) and with a longer follow-up time, an advantage of the bioprostheses in terms of composite event and all-cause mortality would have emerged in the subgroup of older patients. Although selecting only left-sided endocarditis, which is a wise choice considering the important differences in clinical features and long-term prognosis compared with right-sided endocarditis, the Kahrovic study [1] included both patients with native and PVE, which is a source of severely confounding heterogeneity. In a recent study including 395 patients [3], we have observed an advantage of mechanical prostheses in native left-sided endocarditis treatment, with regard to both overall long-term survival and freedom from cardiovascular-related mortality. Despite stricter selection criteria, we were aware of the profound clinical differences between patients receiving a mechanical and a biological substitute, and thus, we balanced the 2 groups by inverse probability of treatment weighting to compare long-term outcomes: thereafter, mechanical valves still resulted superior in terms of all-cause mortality, whereas no significant difference was found for cardiovascular mortality anymore [3]. Of note, even after these methodological precautions, we found that the overall survival advantage of mechanical prostheses was mainly driven by the subgroup of aortic valve replacement, whereas no significant difference was seen in the mitral position [3], suggesting that site of implantation is another source of heterogeneity. This concept can also be applied to the more intriguing question of whether mechanical or biological prostheses are more prone to recurrence of infection when implanted for left-sided endocarditis. Kahrovic et al. [1] found only a non-significant trend in favour of mechanical valves; however, again, we wonder whether confounding factors might have influenced this result. Just to give 2 examples of possible confounders: although the authors relied on direct acyclic graphs, PVE was significantly more prevalent in the group receiving bioprostheses, and abscess exclusion was performed nearly twice more frequently in this group (14.6% vs 7.9%), indicating that perivalvular involvement (a feature increasing significantly the risk of re-endocarditis) was actually more often active or severe in the bioprosthesis group. Would different results have emerged if the authors used different balancing methods (inverse probability of treatment weighting being particularly suited when the numerosity of the compared groups is importantly different)? When confounders are concerned, not only the possibly interfering variables should be identified but also the weight of each of them should be considered. In particular, to discern the influence of PVE inclusion on the observed results, a distinction between prostheses implanted for a previous infective endocarditis (in this case patient characteristics can increase per se the propensity towards further recurrence) and those for other aetiologies would be necessary [4]. Apart from those specific infection-related factors, also host-related factors can be important: why were biological valves chosen in a smaller but still relevant subset of younger patients? Poor life expectancy (e.g. for cancer or other chronic comorbidities) and low socio-cultural level (jeopardizing compliance to long-term anticoagulant therapy) are just 2 examples, being also potential risk factors for recurrence. With this respect, looking at the causes of death in the Austrian study [1], we were impressed by the greater frequency of cancer and of unknown causes of death (at least in our experience associated with a lower socio-cultural level of the patients) in the biological group. In our study [3], with mechanical prostheses, neither the recurrence of endocarditis nor the need for endocarditis-related reoperation differed significantly compared to the outcomes observed after implantation of a biological prosthesis. In conclusion, the debate on the optimal valve substitute for left-sided valve endocarditis is still open and further studies are needed, especially if prospectively designed and with strict inclusion criteria. Before thinking that the retrospectively observed results of valve replacement for infective endocarditis are accounted for by the type of prosthesis, the respective effects of patient selection, criteria for prosthesis choice and analysis methods should be pondered and possibly ruled out.