[Skip to Navigation]
Sign In
Invited Commentary
March 15, 2022

The Never-Ending Quest of Intracerebral Hemorrhage Outcome Prognostication

Author Affiliations
  • 1Department of Neurology, University of California, San Francisco
  • 2Departments of Neurology, Johns Hopkins University, Baltimore, Maryland
  • 3Anesthesiology and Critical Care Medicine, Johns Hopkins University, Baltimore, Maryland
JAMA Netw Open. 2022;5(3):e221108. doi:10.1001/jamanetworkopen.2022.1108

Intracerebral hemorrhage (ICH) remains a challenging disease to treat. Over the past 2 decades, there have been 9 phase III clinical trials testing interventions, such as hemostatic therapies, lowering blood pressure, and surgery for ICH. None have identified a strongly effective treatment, although some headway has been made in reducing mortality. During the same period, there have been at least 60 different published prognostic tools for post-ICH outcomes.1 Most of these tools involve mathematical models or scoring systems developed from relatively modest-sized cohorts of patients from a single or small number of centers, some without external validation, and incorporating generally similar risk factors, including hematoma size, patient clinical condition at ictus, and some other imaging and demographic characteristics. It is into this milieu of frustration and repetition that the Ethnic/Racial Variations of Intracerebral Hemorrhage (ERICH) study brings new insights regarding risk factors associated with ICH outcomes and how to consider their use.

The ERICH study2 is a large multicenter prospective case-control study involving 3000 patients with acute ICH and a similar number of demographically matched controls recruited at 19 different centers in the US. The overall goal of the ERICH study2 is to identify genetic variations associated with risk of ICH in a racially and ethnically balanced population and to determine differences in the distribution of risk factors and imaging characteristics based on race and ethnicity. The ERICH study2 has provided a platform to address numerous questions regarding acute ICH management and outcomes, and the new study by Woo et al3 using the ERICH study cohort reports on risk factors associated with poor outcomes in patients with ICH. In this study,3 76 different potential risk factors were assessed in univariate analysis, and an optimized multivariable model of independent risk factors was assembled using minimization of the Akaike information criterion (AIC). The ERICH study2 enrolled patients with ICH from a preplanned distribution of 1000 White patients, 1000 Black patients, and 1000 Hispanic patients, with race and ethnicity defined by patient or proxy self-report. This study by Woo et al3 of risk factors associated with outcomes included 2568 participants from the ERICH cohort with ICH, as these had available 3-month outcome data on the modified Rankin Scale (mRS), which served as the primary outcome measure, with an mRS score of 3 or less (ambulatory but disabled) considered a good outcome.

What distinguishes the prognostic outcome evaluation by Woo et al3 from other tools is the breadth of parameters assessed. Most published ICH prognostic tools have used parameters assessed at initial patient evaluation as part of regular clinical care. Woo et al3 evaluated factors from multiple different domains, including medical history and comorbidities, hematoma characteristics and other imaging findings, demographics, baseline neurological examination findings, medications, genetics (ie, APOE allele genotype), and numerous subsequent acute in-hospital events or treatments.3 Importantly, the final multivariable model included factors from each of these domains. Common previously identified risk factors associated with outcomes, such as hematoma volume, ICH location, patient score on the Glasgow Coma Scale at hospital admission, intraventricular hemorrhage, age, and significant medical comorbidities (eg, preexisting dementia or disability) were verified as factors associated with the 3-month outcome in the study by Woo et al.3 However, the in-hospital occurrence of infection, need for treatment of elevated intracranial pressure, or tracheostomy placement, as well as preexisting use of α-2 agonist antihypertensive medications, and presence of an APOE 2 allele were also independently associated with 3-month outcomes. It is noteworthy that race and ethnicity as defined in the ERICH study was not independently associated with outcomes.3 The study by Woo et al3 also validated 2 commonly used ICH baseline severity scores, the ICH score and the FUNC score, while demonstrating the overall superiority of their more robust multivariable model compared with these simpler scales. Interestingly and perhaps wisely, Woo et al3 resisted the temptation to propose their own new mathematical model of individual patient’s ICH outcomes based on their findings.

The findings of this study give us the opportunity to reexamine the purpose of clinical grading scales and prognostic models in ICH and to consider their proper and judicious use. The most commonly used and widely validated ICH clinical grading scale, the ICH score, was developed as a simple baseline severity assessment to allow consistency in communication and treatment selection in clinical care and clinical research, but not to attempt to precisely prognosticate outcomes in individual patients.4 It has been interesting to see how this original intent has been shifted toward the desire, and even expectation, of precisely prognosticating patient outcome from a simple scale assessed at ICH onset. It is not that simple, and the study by Woo et al3 reminds us of this. The identification of subsequent events that are targets for intervention, such as prevention of infection, intracranial pressure control, and hematoma expansion, is reassuring that the die is not cast at ICH onset.

So when it comes to identifying the universe of risk factors associated with ICH outcomes, are we finished? Not a chance. Along with exposing the naivete of thinking that a prognostic tool determined at ICH onset sufficiently describes ICH outcome, the study by Woo et al3 also is limited in its assessment of potential risk factors associated with ICH outcomes. Unfortunately, most of our studies of ICH outcome, including the study by Woo et al,3 include limited to no information about the quality and depth of services provided after acute hospital discharge. Yet it is generally believed (perhaps intuitively) that rehabilitation, family support, and prevention of secondary cardiovascular events are important. By not including the postacute care continuum in assessments of outcome prognoses, we are likely also missing targets for intervention. Also, recent reports of patients with acutely disordered consciousness (including those with ICH) have found imaging and electrophysiology markers of activity that are associated with more favorable outcomes.5,6 However, current ICH prognosis tools have almost exclusively relied on information available from existing clinical care, as opposed to investigating the detailed anatomic and functional connectivity causing weakness, impaired consciousness, aphasia, or other clinical findings in specific patients. Ongoing efforts, such as the Curing Coma campaign seek to identify underlying patient endotypes that may be defined by their functional brain networks, genetics or genomics, or other biomarkers.7

Beyond specific independent variables, it is also incumbent on ICH investigators to improve the overall quality of outcome scales. Many simpler prior ICH outcome prognosis models are biased by withdrawal of care (WOC), potentially leading to a self-fulfilling prophecy of poor outcome when using these scores. Of note, Woo et al3 found that their more detailed models, both including and excluding patients with WOC, had similar characteristics. However, one factor to consider is that many patients with worse ICH severity were likely excluded from the ERICH study, and it is these patients who often present the greatest challenge to decision-making regarding goals of care. Similarly, the ERICH cohort had relatively few patients who underwent surgical treatment, and this may be a patient group with a need for targeted prognostication depending on procedural success of the intervention. The optimal timing of use of prognostic scales is further uncharted territory, and the study by Woo et al3 suggests patience, given that ongoing events may influence recovery trajectories. Finally, simply dichotomizing outcome scales, such as the mRS, may not sufficiently represent more nuanced patient-centered outcomes relevant to an acceptable quality of life after ICH.

This study by Woo et al3 assessing risk factors associated with outcomes should be used to define a shift in the way we approach prognosing ICH outcome. We probably do not need more prognostic models that use the same standard parameters but just with slightly different cutoff points or organization. We also need to resist the urge to provide precise outcome prognosis at onset, or probably even early after ICH. We need to hold ICH prognostication to the same standards that we do for our weather forecast or travel times on navigation apps: uncertainty is inherent, but a framework for possible outcomes is still useful. Furthermore, future efforts at prognosing outcome should bring in additional relevant parameters, such as postacute rehabilitation care and advanced assessment of cerebral functional integrity and capacity for recovery. The purpose of providing a prognosis for ICH outcomes should be the same as the purpose of a clinical trial: trying to identify treatments that can help patients. We believe that we are just getting started.

Back to top
Article Information

Published: March 15, 2022. doi:10.1001/jamanetworkopen.2022.1108

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

Corresponding Author: J. Claude Hemphill III, MD, MAS, Department of Neurology, Zuckerberg San Francisco General Hospital, 1001 Potrero Ave, Bldg 1, Rm 101, San Francisco, CA 94110 (claude.hemphill@ucsf.edu).

Conflict of Interest Disclosures: Dr Ziai reported receiving grants from the National Institutes of Health, National Institute of Neurological Disorders and Stroke, and National Institute on Aging and personal fees from C. R. Bard outside the submitted work. No other disclosures were reported.

Gregório  T, Pipa  S, Cavaleiro  P,  et al.  Prognostic models for intracerebral hemorrhage: systematic review and meta-analysis.   BMC Med Res Methodol. 2018;18(1):145. doi:10.1186/s12874-018-0613-8PubMedGoogle ScholarCrossref
Woo  D, Rosand  J, Kidwell  C,  et al.  The Ethnic/Racial Variations of Intracerebral Hemorrhage (ERICH) study protocol.   Stroke. 2013;44(10):e120-e125. doi:10.1161/STROKEAHA.113.002332PubMedGoogle ScholarCrossref
Woo  D, Comeau  ME, Venema  SU,  et al.  Risk factors associated with mortality and neurologic disability after intracerebral hemorrhage in a racially and ethnically diverse cohort.   JAMA Netw Open. 2022;5(3):e221103. doi:10.1001/jamanetworkopen.2022.1103Google Scholar
Hemphill  JC  III, Bonovich  DC, Besmertis  L, Manley  GT, Johnston  SC.  The ICH score: a simple, reliable grading scale for intracerebral hemorrhage.   Stroke. 2001;32(4):891-897. doi:10.1161/01.STR.32.4.891PubMedGoogle ScholarCrossref
Claassen  J, Doyle  K, Matory  A,  et al.  Detection of brain activation in unresponsive patients with acute brain injury.   N Engl J Med. 2019;380(26):2497-2505. doi:10.1056/NEJMoa1812757PubMedGoogle ScholarCrossref
Jang  SH, Chang  CH, Jung  YJ, Kim  JH, Kwon  YH.  Relationship between impaired consciousness and injury of ascending reticular activating system in patients with intracerebral hemorrhage.   Stroke. 2019;50(8):2234-2237. doi:10.1161/STROKEAHA.118.023710PubMedGoogle ScholarCrossref
Provencio  JJ, Hemphill  JC, Claassen  J,  et al; Neurocritical Care Society Curing Coma Campaign.  The Curing Coma campaign: framing initial scientific challenges-proceedings of the first Curing Coma campaign scientific advisory council meeting.   Neurocrit Care. 2020;33(1):1-12. doi:10.1007/s12028-020-01028-9PubMedGoogle ScholarCrossref