See Article by Gulati et al

We demand rigidly defined areas of doubt and uncertainty!

—Douglas Adams, The Hitchhiker’s Guide to the Galaxy

In Douglas Adams’ The Hitchhiker’s Guide to the Galaxy, a group of “hyperintelligent pan-dimensional beings” create an omniscient computer (dubbed Deep Thought) as the ultimate arbiter of truth.¹ The creation of this computer greatly discomfits a pair of philosophers, Vroomfondel and Majikthise, who worry that it will put them “straight out of a job.” As the pair’s vexation grows, Vroomfondel makes his demand to Deep Thought for “rigidly defined areas of doubt and uncertainty” (with the obvious and ridiculous corollary demand for rigidly defined areas of certainty). As physicians, we often feel the same discomfiture—particularly when a study calls into question a concept we had previously assigned as certain or a clinical practice ingrained by biological plausibility. In their important study about the association of preimplant PDE5i (phosphodiesterase-5 inhibitor) use with post-left ventricular assist device (LVAD) outcomes, Gulati et al² have further advanced the boundary of doubt and uncertainty surrounding the use of these medications in patients undergoing LVAD implantation and indeed in any patient with underlying left heart disease (LHD).²

In this issue of Circulation: Heart Failure, Gulati et al retrospectively queried the Interagency Registry for Mechanically Assisted Circulatory Support registry to investigate the association of pre-LVAD PDE5i use and post-LVAD severe early right heart failure (RHF). The primary outcome was defined largely per the current Interagency Registry for Mechanically Assisted Circulatory Support definition: death from RHF or multiorgan system failure within 30 days, need for right ventricular assist device within 30 days, or use of inotropes beyond 14 days.³ The authors excluded patients who received durable right ventricular assist devices during the index procedure from the primary outcome, as it was thought this likely represented a preventive instead of reactive operative strategy. Given the retrospective and nonrandomized nature of the study, selection bias was addressed via a robust propensity matching system using 22 variables in the assigned propensity score to model the likelihood of receiving PDE5i therapy. Importantly, the authors also took pains to perform quality control on the registry-reported hemodynamic data by ensuring pulmonary arterial wedge pressure did not exceed pulmonary arterial diastolic pressure, carefully selecting between right atrial and central venous pressures based on plausibility, and not accepting pulmonary arterial wedge pressures <1 mm Hg.

Ultimately, the study compared 1177 propensity-matched PDE5i and control patients. Even with the extensive propensity matching model, severe early RHF remained more common in the PDE5i group versus controls (odds ratio, 1.31; 95% CI, 1.09–1.57). Interestingly, this was driven solely by prolonged inotrope duration, with no significant difference in right ventricular assist device use or death from RHF. Furthermore, although total hospital length of stay was 3 days longer in the PDE5i group, at 6 months the groups had similar survival, quality of life scores, and 6-minute walk distance. The authors astutely investigated the association of PDE5i use and major bleeding events, finding a significantly higher incidence of major bleeding within the first 7 postoperative days in the PDE5i group (hazards ratio, 1.52; 95% CI, 1.15–2.0).

The incautious interpretation of the study is that preoperative PDE5i use increases the risk for post-LVAD severe early RHF. However, several aspects of the study counsel caution in reaching this conclusion. First, despite extensive propensity matching endeavors, it is possible that residual selection bias exists in this retrospective study—it may still be that patients receiving preoperative PDE5i therapy were sicker in unknown unknown ways than the matched control cohort. Notably, of the 22 variables used in the propensity scoring system, pulmonary arterial wedge pressure was missing in 30% of cases. The authors report that pulmonary vascular resistance was higher in the PDE5i group (2.6 versus 2.3 WU, P<0.001), but the high proportion of missing pulmonary arterial wedge pressure values may lead us to underestimate the true difference in pulmonary vascular load between the groups. Second, it is noteworthy that though the authors report a higher incidence of severe early RHF in the PDE5i group, this was driven entirely by prolonged inotrope use. The PDE5i group did not suffer higher early mortality, need for early right ventricular assist device implantation, or even higher morbidity or mortality at 6 months. Many prior studies have found an association between severe early RHF and mortality,^4–6 and so its absence here is conspicuous and leads us to question exactly what specific outcome we are capturing—true hemodynamic failure of the right heart system, covariation of preoperative PDE5i use and postoperative provider decision making regarding inotrope use, or association of PDE5i use with a postoperative event necessitating prolonged inotrope use but not representing causative hemodynamic failure of the right heart? The finding of a higher risk of early postoperative bleeding in the PDE5i group points a tentative finger at the latter explanation. However, we must note that the propensity scoring system did not account for important bleeding risk factors, such as platelet count and international normalized ratio. It is possible that the noted bleeding risk was itself confounded by unaccounted for selection bias.

Despite these caveats, this study adds an important voice to others which together push the use of PDE5i therapy in patients with pulmonary hypertension due to LHD into an increasingly “rigidly defined area of doubt and uncertainty.” It is notable that, of the patients meeting the inclusion criteria in this study, 10.4% were on preoperative PDE5i therapy—a striking proportion in a patient population selected for the most severe degree of LV failure and ostensibly in whom preoperative right ventricular embarrassment is selectively avoided. Prior studies have indicated that the majority of real-world PDE5i use is in patients without WHO Group 1 precapillary pulmonary arterial hypertension.^7,8 In one study of veterans treated with PDE5i, over half did not even have a verifiable right heart catheterization.⁷ We might be forgiven to a degree for this prescribing pattern; we know that patients with pulmonary hypertension due to LHD do worse, and as clinicians we are desperate for helpful therapies in this scenario. The afterload sensitive nature of the right ventricle,⁹ the ability of PDE5i therapy to reduce pulmonary vascular load,¹⁰ and early small studies with promising surrogate end points¹¹ all present an enticing, biologically plausible argument for their use in these patients. However, together with the recently published SIOVAC study (Sildenafil for Improving Outcomes After Valvular Correction) showing worse outcomes with PDE5i therapy in patients with persistent pulmonary hypertension after valvular intervention,¹² the current study weakens the arguments of biological plausibility for the utility of these agents in patients with LHD.

Importantly, this study begs the important next question of the role of PDE5i therapy initiated as a treatment for RHF after LVAD implantation. Other important areas of doubt and uncertainty urgently requiring answer include whether any patient phenotype with pulmonary hypertension due to LHD benefits from PDE5i therapy and whether other pulmonary arterial hypertension–specific therapy may be more effective for these patients. The currently enrolling SOPRANO study may shed light on the utility of endothelin receptor antagonist therapy post-LVAD,¹³ although doubtless questions will remain. In the current study, Gulati et al are to be congratulated not for providing an answer to these questions but for more rigidly defining the area of doubt and uncertainty. In doing so, they have provided what Lewandowsky et al¹⁴ describe as uncertainty as actionable knowledge. The call to action as presented by this study is increasingly pressing—if pulmonary arterial hypertension–specific therapy is indeed shown to increase the risk of harm to our patients with LVAD or LHD, guidelines and practice patterns should expeditiously abandon the arguments of biological plausibility and surrogate end points just as they did with antiarrhythmic agents for premature ventricular contractions and inotropic medications for LV failure.

Ultimately, the philosophers in The Hitchhiker’s Guide are placated when Deep Thought assures them that it will take 7.5 million years to reach "the answer to the Ultimate Question of Life, the Universe, and Everything," giving them plenty of time to ply their trade. Hopefully, we will not have to wait that long to answer these important questions regarding pulmonary arterial hypertension–specific therapy in patients with LVAD or LHD.

None.

The opinions expressed in this article are not necessarily those of the editors or of the American Heart Association.

Guest Editor for this article was Kenneth B. Margulies, MD.

参见Gulati等人的文章

我们要求严格定义怀疑和不确定的领域！

—道格拉斯·亚当斯（Douglas Adams），《银河漫游指南》

在道格拉斯·亚当斯（Douglas Adams）的《银河漫游者指南》中，一群“超智能的超尺寸生物”创造了一种无所不知的计算机（称为“深层思想”）作为真理的终极仲裁者。^1个这台计算机的创建极大地使一对哲学家Vroomfondel和Majikthise感到不适，他们担心这会使他们“直接失业”。随着两人烦恼的加剧，Vroomfondel向“深度思考”提出了对“严格定义的不确定性和不确定性区域”的要求（对严格确定性的确定性区域提出了明显而荒谬的必然要求）。作为医生，我们经常会感到同样的困惑-尤其是当一项研究提出疑问时，我们先前曾将我们先前指定为某些概念或因生物学上的可行性而根深蒂固的临床实践提出了疑问。在他们关于植入前PDE5i（磷酸二酯酶5抑制剂）的使用与左心室辅助装置（LVAD）结局的关联的重要研究中，Gulati等[ ^2]LVAD植入患者以及任何潜在的左心疾病（LHD）患者使用这些药物的前景已进一步受到质疑和不确定性的界限。^2个

在本期《循环：心力衰竭》中，Gulati等人回顾性询问了机械辅助循环支持登记间机构注册表，以调查LVAD之前使用PDE5i与LVAD后严重早期右心衰竭（RHF）的关联。主要结果主要根据当前的机械辅助循环支持机构间注册中心定义：30天内因RHF或多器官系统衰竭而死亡，30天内需要右心室辅助装置或14天内使用正性肌力药。³作者从主要结局中排除了在索引过程中接受耐用右心室辅助装置治疗的患者，因为认为这很可能是预防性的方法，而不是反应性的手术策略。考虑到研究的回顾性和非随机性，选择偏倚可以通过强大的倾向匹配系统解决，该系统使用分配的倾向评分中的22个变量来模拟接受PDE5i治疗的可能性。重要的是，作者还努力确保注册机构报告的血流动力学数据达到质量控制，方法是确保肺动脉楔压不超过肺动脉舒张压，并根据合理性在右心房和中心静脉压之间谨慎选择，并且不接受肺动脉楔形压力<1毫米汞柱。

最终，该研究比较了1177名倾向匹配的PDE5i和对照患者。即使采用广泛的倾向匹配模型，PDE5i组与对照组相比，严重的早期RHF仍较常见（优势比，1.31； 95％CI，1.09-1.57）。有趣的是，这仅是由于延长了药物的持续时间导致的，而右心室辅助装置的使用或RHF的死亡没有显着差异。此外，尽管PDE5i组的总住院时间长了3天，但在6个月时，这些组的生存率，生活质量得分和步行距离都只有6分钟。作者对PDE5i的使用与严重出血事件之间的关系进行了深入研究，发现PDE5i组术后7天内严重出血的发生率显着更高（危险比，1.52； 95％CI，1.15-2.0）。

对研究的不正确解释是，术前使用PDE5i会增加LVAD后发生严重早期RHF的风险。但是，研究顾问的多个方面在得出此结论时要谨慎行事。首先，尽管进行了广泛的倾向匹配努力，但这项回顾性研究仍可能存在残留的选择偏见-仍可能是接受术前PDE5i治疗的患者比匹配的对照队列病情未知。值得注意的是，在倾向评分系统中使用的22个变量中，有30％的病例缺少肺动脉楔压。作者报告说，PDE5i组的肺血管阻力较高（2.6 vs 2.3 WU，P<0.001），但是肺动脉楔压值缺失的比例很高，可能导致我们低估了两组之间的肺血管负荷的真正差异。其次，值得注意的是，尽管作者报告说PDE5i组中严重的早期RHF发生率较高，但这完全是由于长期使用了Inotrope所致。PDE5i组没有出现更高的早期死亡率，也没有需要早期右心室辅助装置植入，也没有出现6个月时更高的发病率或死亡率。许多先前的研究发现严重的早期RHF与死亡率之间存在关联，^4-6因此，它的缺席是显而易见的，并导致我们确切地质疑我们正在捕获什么具体结果-右心脏系统的真正血液动力学衰竭，术前PDE5i使用的协变和术后治疗者对药物的使用的决策，或PDE5i与APO的关联。术后事件是否需要长期使用Inotrope，但不能代表右心脏的因果性血流动力学衰竭？在PDE5i组中发现较高的术后早期出血风险是用手指指着后面的解释。但是，我们必须注意，倾向评分系统并未考虑重要的出血风险因素，例如血小板计数和国际标准化比率。所指出的出血风险本身可能是由于无法解释选择偏见而造成的。

尽管有这些警告，但这项研究为其他人增加了重要的声音，这些人将PDE5i治疗在LHD引起的肺动脉高压患者中的使用推向越来越严格的“不确定性和不确定性领域”。值得注意的是，在符合本研究纳入标准的患者中，有10.4％的患者接受了术前PDE5i治疗，这一比例在选择最严重的LV衰竭程度且表面上选择性地接受术前右心室窘迫的患者人群中占了惊人的比例。避免。先前的研究表明，现实世界中大多数使用PDE5i的患者均没有WHO第1组毛细血管前肺动脉高压。^7,8在一项对接受PDE5i治疗的退伍军人的研究中，超过一半的人甚至没有可验证的右心导管检查。⁷我们可能会在某种程度上原谅这种处方模式。我们知道，由于LHD而导致的肺动脉高压患者的病情会恶化，作为临床医生，我们迫切希望在这种情况下提供有用的治疗方法。右心室的后负荷敏感特性，⁹ PDE5i治疗降低肺血管负荷的能力，¹⁰和早期早期研究（具有有希望的替代终点）¹¹都为这些患者的使用提供了诱人的，生物学上合理的论据。但是，与最近发表的SIOVAC研究（西地那非用于改善瓣膜矫正后的结果）一起显示，PDE5i治疗在瓣膜介入治疗后持续性肺动脉高压患者中的预后较差，[ ^12]当前的研究削弱了这些药物在LHD患者中的效用的生物学可行性的论点。

重要的是，这项研究提出了一个重要的下一个问题，即在LVAD植入后开始作为RHF治疗的PDE5i治疗的作用。迫切需要回答的其他重要疑问和不确定性领域包括，由于LHD导致的任何患有肺动脉高压的患者表型是否都可以从PDE5i治疗中获益，以及其他针对肺动脉高压的疗法是否可能对这些患者更有效。目前正在进行的SOPRANO研究可能会揭示LVAD后内皮素受体拮抗剂治疗的效用[ ^13]，尽管无疑会有疑问。在当前的研究中，不应祝贺Gulati等人提供了这些问题的答案，而是更严格地定义了疑问和不确定性的范围。通过这样做，他们提供了Lewandowsky等¹⁴将不确定性描述为可操作的知识。这项研究提出的行动呼吁越来越迫切—如果确实显示肺动脉高压特异性疗法确实增加了对我们的LVAD或LHD患者的伤害风险，则指南和实践模式应尽快放弃生物学上的合理性和就像使用抗心律不齐药物治疗早发性心室收缩和使用正性肌力药物治疗左室衰竭一样，替代终点。

最终，《思想旅行者》指南中的哲学家被深思熟虑地向他们保证，要花750万年才能达到“生命，宇宙和万物的终极问题的答案”，这给了他们充裕的时间进行交易。希望我们不必等那么久就回答有关LVAD或LHD患者的肺动脉高压特异性治疗的这些重要问题。

没有。

本文表达的观点不一定是编辑者或美国心脏协会的观点。

本文的客座编辑是医学博士Kenneth B. Margulies。