Hello, everyone. I'm Yvella from Children's Hospital Los Angeles. I appreciate the opportunity to speak at SCCM, and I've been asked to comment on the next steps after the Second Pallet Conference and how we move forward towards Parts 3.0. I have no relevant disclosures. The learning objectives that I have for this talk are to describe the knowledge gaps that exist in pediatric ARDS research, to highlight potential approaches to improve the generation of new knowledge in pediatric ARDS, and then lastly, to highlight some potential innovative approaches to clinical trials to improve generation of new knowledge in pediatric ARDS. So, as Dr. Kamani has highlighted, Pallet 2 created 146 recommendations and statements for pediatric ARDS. Of these, there were 34 clinical practice recommendations and 112 consensus-based statements. The clinical practice recommendations were based on evidence directly related to the developed patient intervention, comparator, and outcome, or PICO, questions. Evidence used to develop the clinical practice recommendations included a combination of observational studies, randomized control trials, and meta-analyses in adults and children with a hierarchy of evidence shown to the left. About a third of the clinical practice recommendations were focused on addressing management of mechanical ventilation in pediatric ARDS, where there was the largest amount of evidence to support recommendations. The 34 clinical recommendations from Pallet were developed and graded using the GRADE framework. The good practice statements were also based on evidence, but not evidence supporting an answer to the direct PICO question raised, but rather evidence thought to be generally applicable to the question. The GRADE framework has four levels based on the certainty of evidence supporting a recommendation. Almost all of the Pallet clinical recommendations were graded as low or very low certainty of evidence to support the recommendation. The only recommendation that was graded as moderate or having a moderate confidence in the effect estimate was the mechanical ventilation recommendation to maintain PEEP levels at or above the lower PEEP higher FIO2 table from the ARDS network protocol. To further emphasize this point, of the 34 clinical practice recommendations, the majority were based on either observational data or data applied from adult studies, and only about a third were based on randomized control trial data, with only seven of these being randomized control data from children with PARDS. Even of these seven randomized control trials in PARDS, many of them were single center studies, which of course raises concerns regarding the generalizability of findings to other centers. While we feel that the Pallet recommendations are based on the best evidence that we have available, I hope that this overview highlights the great need to generate further evidence in PARDS in order to continue to improve and expand on our recommendations for management in PARDS 3.0. Studies in PARDS are historically difficult, though. This was highlighted by the prone positioning study that was led by Martha Curley in the early 2000s. This study was limited to children with the AECC definition of ARDS, which we know recognizes fewer patients with PARDS and the Pallet definition of PARDS. But the study still highlights the difficulty that can occur with the recruitment of children with PARDS into clinical studies. This study screened about 8,000 mechanically ventilated patients over a period of three years from seven PICUs. Ultimately, only 184 patients were identified as having acute lung injury, with about 55 percent of these families consented and enrolled in the study. You can see that similar to most clinical trials, this study had many exclusion criteria that are fairly common in children with PARDS, including persistent hypotension and having had a bone marrow transplant, which then further contributes to difficulty with enrollment and also subsequently limits generalizability of findings to broader populations of children with PARDS. While we have very few clinical trials in pediatric ARDS, we know from adult literature that the majority of interventional clinical trials in adults with ARDS have not led to new therapies. In fact, in an analysis several years ago of all clinical trials in adults with ARDS, when limited to larger studies with more than 50 deaths in any one treatment arm, only two clinical trials out of 21 found a benefit. All of these clinical trials had strong preclinical or phase one or two clinical trial data to support them being conducted. However, there is often this lack of translatability that occurs when questions are tested in the setting of randomized controlled trials. There are numerous reasons for this, including truly ineffective interventions or choosing unresponsive patients for the interventions or clinical trial design choices for sample size or endpoints that may impact the feasibility of the trial. On the other hand, there is data to suggest that well-conducted large observational studies may be just as good as randomized controlled trials in some situations. This is a meta-analysis of reviews comparing the results from randomized controlled trials to observational studies when assessing the same hypothesis. You can see that for the vast majority of studies, there is no significant difference in effect size estimates with, if anything, a slight favoring of a higher effect size in the randomized controlled trials and the observational studies. So when we think about trying to develop and further refine recommendations for pediatric ARDS management or move towards PARDS 3.0, we really need to generate additional high-quality data that's specific to children. However, when considering this, we need to think about some of the challenges that exist to generating randomized controlled trial data in PARDS. I already highlighted challenges with enrollment that can occur with only about 3% of PICU admissions being PARDS admissions. Multi-center studies are really required. However, we are also challenged by other factors. As Brian Briscoe spoke about, PARDS is a heterogeneous disease process with different sub-phenotypes and endotypes that may respond differently to the same treatment. Carolyn Kelsey has described the inflammatory and hypo-inflammatory phenotypes of ARDS, which have also been recently found in children, with the different phenotypes having significantly different mortality risk and also evidence for a differential response to interventions such as fluid management in PEEP or medications such as simvastatin. Studies focused on a particular endotype or sub-phenotype, particularly when numerous exclusion criteria exist, however, then struggle from potential lack of generalizability to the larger cohort of pediatric ARDS. Other challenges that are shared with most other clinical trials in critical care are time-sensitive enrollment. For pediatric ARDS, the first 24 hours is a time period of rapid change. Interventions may be most effective during early pediatric ARDS. However, consent and enrollment in clinical trials is often difficult during this time period. Furthermore, how do we think about studying interventions that are commonly used but may be rescue therapies in pediatric ARDS, such as inhaled nitric oxide? There may not be equipoise to randomize patients in some situations. Additionally, thoughtful selection of the appropriate outcome is crucial to well-designed clinical trials in pediatric ARDS. It's important to choose outcomes that are patient-oriented and meaningful. They need to be easily and reliably measured, but they also obviously need to be plausibly changed by the intervention that's being tested. While long-term outcomes might be most important to families and patients, there are significant challenges that exist to following patients after hospital discharge. And lastly, for all these reasons, funding is really necessary for most clinical trials. However, funding is becoming increasingly more difficult to obtain. So, what are some methods to try and address these challenges to generating the needed additional data to guide our future Parts 3.0 definitions and recommendations? I would really advocate for two main approaches that require multicenter collaboration. First, we need to commit to multicenter observational cohort studies in pediatric ARDS. And two, we need to consider innovative clinical trial approaches and designs that make multicenter randomized clinical trials more efficient and feasible for children. This is particularly important in pediatric ARDS given the much smaller numbers of patients that we have available to study. Although randomized control trials are really the gold standard for testing the efficacy of an intervention, observational studies are less expensive to conduct and may produce results more quickly, although they are potentially less reliable than a randomized control trial. Furthermore, all interventions may not require the same level of evidence to change clinical practice. For example, an intervention with minimal risk may have sufficient evidence from an observational trial to consider changing practice. One of the challenges to studying ARDS in existing data cohorts is that ARDS is poorly recognized clinically. There's evidence to suggest that up to 40% of patients with ARDS are never recognized as having ARDS at the bedside. This is why PALIC-2 developed a recommendation to use electronic tools to automatically screen and help identify children with pediatric ARDS. However, this is not yet mainstream at most institutions. Therefore, existing data sets are not practical for understanding phenotypes within ARDS. This leads to multicenter prospective observational cohort studies with highly granular clinical, physiologic, radiologic, biological data and samples where centers adhere to best practice as really crucial to test hypotheses with observational data. Studies such as this would benefit from integrated data capture to improve high-quality data collection. They could allow for deep phenotyping of ARDS patients with the ability to identify, validate, and refine clusters of subphenotypes and factors that are associated with biological differences or differences in the treatment response interventions and parts. High-quality observational data may be our best method of estimating the effect of some rescue therapies as well where randomized clinical trials may not be possible. Additionally, observational data can also be used to assess the generalizability of results from interventional trials to more general cohorts of pediatric ARDS. In March 2021, the NHLBI announced a research initiative to comprehensively and prospectively phenotype patients with ARDS, pneumonia, and sepsis, the APS consortium. Through this consortium, the biological imaging and clinical characteristics of 5,000 adult patients with critical illness will be assessed for the year of follow-up to study the longer-term effects of ARDS. Similar efforts in children would be highly beneficial. The PARTY study was this type of prospective granular observational study. However, it lacked biological samples. PARTY did enroll 708 children with pediatric ARDS from 145 international PICUs over 10 distinct study weeks from 2016 to 2017. The observational data from this study has generated seven published manuscripts to date, all of which have been used to guide the PALIC-2 recommendations in some way. The manuscripts highlight the impact of studies such as this on moving our understanding of PARDS forward. Given that most clinical trials of ARDS have not led to new therapies and pediatric ARDS has limited numbers of patients to study, there is also a significant need to improve clinical trial design to improve the likelihood of identifying beneficial interventions. Clinical trial design can be optimized in many ways to improve our likelihood of finding meaningful data to support PARDS management. We can optimize patient selection either through predictive or prognostic enrichment. Thoughtful selection of primary and secondary endpoints is important with the appropriate statistical approaches to analyze outcomes, particularly those with skewed distributions. Strategies for conducting the trial such as integrated data capture may improve clinical trial performance. We can also consider innovative clinical trial designs, such as adaptive clinical trial designs, platform clinical trials, and pragmatic approaches to hypothesis testing. All of these methods can in some way make clinical trials more efficient, feasible, or generalizable, which are all important in pediatric ARDS. So classically clinical trials in ARDS have sought to have fairly broad inclusion criteria to facilitate enrollment. This type of approach helps the generalizability to the general ARDS population but has the downside of requiring large sample sizes and potentially missing subgroups of patients that may have differences in response to treatments, or some subjects may respond to a particular intervention with benefit while others are harmed or have no benefit. Given ARDS is a heterogeneous clinical syndrome, this is particularly relevant to ARDS research. Enrichment of clinical trials makes them more efficient with smaller sample sizes required. Prognostic enrichment seeks to increase the likelihood that the outcome of interest occurs in the cohort. Predictive enrichment is more specific and was when you increase the likelihood patients in the trial have a therapeutic effect from the tested intervention. Both of these methods make the clinical trial more feasible with a smaller sample size but also predictive enrichment allows for more directed interventions, potentially the biological or phenotypic features of pediatric ARDS. Adaptive clinical trials can also improve the efficiency and feasibility of a clinical trial. An adaptive clinical trial uses interim data analysis to guide modifications of the ongoing trial without undermining the validity or the integrity of the study. Potential modifications are considered in advance and are pre-specified. The increased use of Bayesian statistical methods have made adaptive trial designs more feasible. With the Bayesian approach, assumptions about the relationship between an intervention and the outcome are made then reassessed as data is collected with revisions of the original assumptions as the trial continues. Advantages to adaptive clinical trials are that the sample size can be refined so the trial could be stopped early or modified depending on the results of interim analyses. Assignment arms of the trial can also be adapted as the trial is ongoing to select arms or doses that are demonstrating more benefit for higher enrollment or even to remove arms that are demonstrating minimal benefit. This can help trials be more efficient, informative, and ethical. The currently enrolling prone and oscillation pediatric clinical trial study or prospect is an example of an adaptive randomized controlled clinical trial that is currently enrolling in children with ARDS. The study is enriched with patients most likely to benefit from lung recruitment, those with moderate to severe ARDS, and patients are randomized to test the hypothesis that prone versus supine and high frequency ventilation versus conventional ventilation will improve ventilator-free days. Assignment to study arms will be adapted using data collected through the trial. Platform clinical trials are another avenue to improve clinical trial efficiency and feasibility that may be useful for pediatric ARDS. Platform clinical trials are becoming increasingly common for several reasons. Basically, a platform clinical trial is an ongoing clinical trial that has a master protocol. Through this type of clinical trial, multiple interventions can be tested against a common control. Therapies can be added or removed as enrollment is ongoing. Familiarity and consistency with a master protocol makes it easier to conduct multiple randomized controlled trials within the platform. This makes studying interventions more time-efficient and cost-effective. The COVID-19 pandemic highlighted the benefit of these types of trials. This graphic demonstrates the platform clinical trials in COVID-19 that were being conducted in just the first two years of the pandemic. Many of these trials gave us early important information about how to manage and treat COVID-19-related ARDS. Some of these trials utilized novel approaches, such as automatic data capture from the EMR and hiring off-site data entry specialists to deal with staffing shortages at individual sites. So, in summary, I hope I have demonstrated to you that there are substantial knowledge gaps that exist in PARDS that we urgently need to address to move the care and management of children with PARDS forward and develop PARDS 3.0. Collaborative efforts, either through multicenter observational cohort studies or multicenter clinical trials, are required to generate the necessary data required. There are newer methods for clinical trial design that may overcome some of the challenges we have in studying pediatric ARDS and help optimize the likelihood that, over the next several years, we will generate new data that's required to move PARDS towards PARDS 3.0. I'll just highlight here in closing that, in 2015, the PALICC 1 recommendations came out, and you can see the growth in research from this PubMed graph that occurred after the recommendations were published. So, I hope that PALICC 2 similarly launches a plethora of new interest in research in pediatric ARDS to develop recommendations for PARDS 3.0.

Pediatric Acute Respiratory Distress Syndrome

knowledge gaps

clinical practice recommendations

observational data

randomized controlled trials