[Skip to Navigation]
Sign In
Invited Commentary
March 10, 2022

ADAPTing to a New Era of Comparative Effectiveness Research in Traumatic Brain Injury—Generating Evidence From Observational Data

Author Affiliations
  • 1Department of Neurological Surgery, University of California, San Francisco
  • 2Center for Medical Decision Sciences, Erasmus MC, University Medical Center, Rotterdam, the Netherlands
  • 3Department of Neurosurgery, Antwerp University Hospital and University of Antwerp, Edegem, Belgium
JAMA Netw Open. 2022;5(3):e220899. doi:10.1001/jamanetworkopen.2022.0899

Abnormalities in brain water balance play an important role in the pathophysiology of traumatic brain injury (TBI) and other neurologic disorders. Brain edema, the result of an abnormal increase in brain tissue water content, is associated with increased intracranial pressure (ICP) that can potentially lead to brain ischemia, cerebral herniation, and death. Despite the clinical importance of brain edema, the molecular mechanisms of brain water balance remain poorly understood, and current treatments have changed little since the introduction of intravenous injections of hypertonic solutions (30% sodium chloride) by Weed and McKibben1 more than 100 years ago. The most commonly used hypertonic solutions for the treatment of elevated ICP are mannitol and hypertonic saline (HTS). However, data are limited on the most effective doses of these agents, and debate remains ongoing over which hypertonic solution is superior. In the pediatric population, HTS (3%) is a level 2 recommendation in the recent guidelines, but this recommendation is based on studies with few patients.2 This important clinical question was addressed by the Approaches and Decisions for Acute Pediatric TBI (ADAPT) investigators, who conducted a large, multicenter, observational cohort study in which they characterized current use of HTS and mannitol and compared their effectiveness.2

This study should be seen as comparative effectiveness research (CER) in which effects of HTS and mannitol on ICP and cerebral perfusion pressure are compared. It illustrates the opportunities provided by large-scale collaborative observational studies but also the challenges involved in CER using nonrandomized studies. In randomized studies, the treatment allocation is random, which makes treatment and control patients by definition comparable, and any difference in outcome can be attributed to the treatment under study. In observational studies, the treatment allocation is clearly not random. The challenge is to assess whether the treatment intervention was targeted appropriately and to understand the treatment allocation mechanism to be able to derive an unbiased treatment effect. In this study, hyperosmolar treatment was targeted to patients with increased ICP and not specifically to those with cerebral edema. Increased ICP can also be caused by brain swelling from vasodilation, in particular in pediatric populations. It would appear unlikely that hyperosmolar agents would exert a mechanistic effect in patients with a vascular cause of increased ICP. As a consequence, treatment effects may be diluted. Treatment allocation is typically based on patient characteristics; for example, in observational studies exploring the effect of ICP monitoring in outcome, the sickest patients will receive the monitoring; therefore, outcome may well be poorer. In the current study, the authors handle such confounding by indication by adjusting for differences in baseline characteristics and for the level of cerebral perfusion pressure and ICP just before bolus administration through covariate adjustment and propensity scores. This approach results in an unbiased treatment effect estimate under the assumption that all of the treatment allocation mechanism is fully captured in the adjustment variables (ie, that all factors that may differ between the treatment groups are known and measured and included in the analysis). However, this assumption cannot be tested and might have been violated in this study. Examples of possible residual confounders include the serum osmolarity before the bolus administration, previously administrated doses, and the rationale for hyperosmolar therapy (eg, given prophylactically as therapy for demonstrated increased ICP or as part of longer-duration therapy to control ICP after an episode of increased ICP). Furthermore, HTS may in some cases have been chosen as the preferred agent in patients with major extracranial injuries, taking benefit of volume expansion. Even in the best-case scenario, it would be virtually impossible to predefine and collect all possible confounders. An additional challenge in studies on intensive care unit therapies in TBI is time-varying confounding; osmolar therapies are repeatedly given based on, for example, ICP, which means that ICP on time 1 (T1) is a confounder for osmolar therapy at time 1, subsequent ICP measurement on time 2 is on the causal pathway for the effect of osmolar therapy on time 1 but a confounder for osmolar therapy on time 2, and so on. The authors decided to focus on the measurement before and after bolus administration in the analysis to avoid time-varying confounding. Alternatively, they could have used causal approaches, such as marginal structural models,3 but these methods are complicated and not yet used in TBI research. Another alternative that has been used in TBI research could have been to exploit the substantial practice variation that was observed by performing instrumental variable analysis.4 However, instrumental variable analysis also relies on assumptions, and statistical power is relatively low.

The authors have done the best they could under the conditions of an observational study with multiple aims and should be congratulated for their extensive and state-of-the-art approaches to adjust for such confounding. Nevertheless, it is crucial to realize that treatment effect estimates from observational CER studies are valid only if the untestable assumptions (eg, no unmeasured confounding) that are made in the analysis are true. Ideally, one uses multiple methods with different assumptions to test the robustness of the results. Interpreting the results of CER requires understanding of the treatment allocation mechanisms and understanding of the methods—in other words, close collaboration between clinicians and epidemiologists or statisticians, as carried out by the ADAPT team.

This large, multicenter study provides real-world evidence, in conjunction with the current pediatric TBI guidelines, that supports the use of HTS as a first-line agent in the management of pediatric TBI. This study also highlights the importance of large-scale observational studies and the evolving importance of CER and other novel clinical trial designs. Randomized clinical trials, as currently conducted, have become more expensive and time consuming. Furthermore, many randomized clinical trials exclude clinically relevant patients, which diminishes the relevance of the trial results to a broader population. This limitation is particularly true for TBI for which some trials have failed to change clinical practice. Given the substantial sample size of ADAPT and other recent large-scale observational TBI studies, such as CENTER-TBI and TRACK-TBI, there are many opportunities ahead to document clinical practice and evaluate the effectiveness of our current interventions.

Back to top
Article Information

Published: March 10, 2022. doi:10.1001/jamanetworkopen.2022.0899

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Manley GT et al. JAMA Network Open.

Corresponding Author: Geoffrey T. Manley, MD, PhD, Department of Neurological Surgery, University of California, San Francisco, 1001 Potrero Ave, Bldg 1, Room 101, San Francisco, CA 94110 (manleyg@ucsf.edu).

Conflict of Interest Disclosures: Dr Maas reported receiving grants for CENTER-TBI from the Framework 7 programme of the European Union, the Hannelore Kohl Stiftung, Integra LifeSciences, and NeuroTrauma Sciences. Dr Lingsma reported receiving grants for CENTER-TBI from the Framework 7 programme of the European Union. Dr Manley reported receiving grants from the US Department of Defense TBI Endpoints Development Initiative and the US Department of Defense TRACK-TBI Precision Medicine, the US Department of Defense/Medical Technology Enterprise Consortium TRACK-TBI Network, and the National Institute of Neurological Disorders and Stroke TRACK-TBI, support for a precision medicine collaboration from the US Department of Energy Support, TRACK-TBI patients stipends and support to clinical sites from One Mind Funding, and support for TRACK-TBI data curation efforts from NeuroTrauma Sciences LLC during the conduct of the study and grants from National Football League Scientific Advisory Board outside the submitted work. No other disclosures were reported.

Weed  LH, McKibben  PS,.  Experimental alteration of brain bulk.   Am J Physiol. 1919;48:531-558. doi:10.1152/ajplegacy.1919.48.4.531 Google ScholarCrossref
Kochanek  PM, Adelson  PD, Rosario  BL,  et al; ADAPT Investigators.  Comparison of intracranial pressure measurements before and after hypertonic saline or mannitol treatment in children with severe traumatic brain injury.   JAMA Netw Open. 2022;5(3):e220891. doi:10.1001/jamanetworkopen.2022.0891Google Scholar
Robins  JM, Hernán  MA, Brumback  B.  Marginal structural models and causal inference in epidemiology.   Epidemiology. 2000;11(5):550-560. doi:10.1097/00001648-200009000-00011 PubMedGoogle ScholarCrossref
Cnossen  MC, van Essen  TA, Ceyisakar  IE,  et al.  Adjusting for confounding by indication in observational studies: a case study in traumatic brain injury.   Clin Epidemiol. 2018;10:841-852. doi:10.2147/CLEP.S154500 PubMedGoogle ScholarCrossref