Medical futility regarding cardiopulmonary resuscitation in in-hospital cardiac arrests of adult patients: A systematic review and Meta-analysis

,


Background
2][3][4] Additionally, a considerable proportion of CPR survivors suffer from subsequent neurologic disabilities. 2,4,5Despite advances in critical care and resuscitation measures, survival rates have only slightly increased during the last decades [5][6][7] and even decreased in some patient groups, such as elderly patients. 3While CPR was originally intended for patients who experience a sudden and unexpected cardiac arrest with a presumed high chance of functional recovery, 8 the intervention has become a standard procedure performed in almost any cases of cardiac arrests. 9However, in hospitalized patients with severe illness and/or debilitating comorbidities survival to discharge with a favorable neurologic outcome is highly unlikely.In such patients, CPR may be considered medically futile. 1,10Yet, in clinical practice, there are no established criteria to determine medical futility regarding resuscitation in case of IHCA.

Importance
The concept of medical futility regarding resuscitation has been discussed for decades with no international consensus being achieved.Common criticisms concern the usefulness and implementation of futility in clinical practice due to a lack of established criteria and ethical considerations. 11,12Still, futility remains an essential topic for clinical decision-making in daily practice. 1,13There are national guidelines such as the medical-ethical guidelines on code status decisions published by the Swiss Academy of Medical Sciences. 14hese state that resuscitation is indicated if there is a chance that the patient survives without severe neurologic impairments.However, because definitions of futility are rather vague and lack specific criteria, implementation in clinical practice remains difficult.There have been different attempts to define futility based solely on expected in-hospital mortality rates without considering neurological outcome. 15However, accurate estimation of survival after CPR is challenging and varies significantly among different physicians. 16In one study, physicians have overestimated the likelihood of survival of adult patients with IHCA by as much as 300%, 17 while another study found that physicians can predict patient survival after IHCA no better than chance. 18Further, a significant number of patients with a very low likelihood of survival after IHCA still have no donot-resuscitate (DNR) orders in place in clinical practice. 19

Goals of this investigation
The objective of this systematic review and meta-analysis was to understand the concept of medical futility regarding CPR in case of IHCA of adult patients by investigating definitions of medical futility regarding resuscitation, assessment of futility for in-hospital CPR, and the prevalence of DNR orders in hospitalized patients in whom CPR would be deemed futile.

Methods
Types of studies, participants, and outcomes Data collection and reporting followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. 20We included peer-reviewed studies discussing and/or evaluating medical futility regarding CPR in adult hospitalized patients.Studies were eligible if they reported either a definition of futility regarding CPR, clinical measures to assess futility, and/or rates of DNR orders in patients for whom CPR attempt was deemed futile.No restrictions concerning age or gender of adult participants were applied.No publication date restrictions and no language restrictions were used.
Studies were excluded if one of the following criteria was present: 1) medical futility regarding resuscitation not addressed / population does not include patients in whom futility regarding resuscitation is assessed, 2) patients < 18 years, 3) no clinical peer-reviewed study or conference poster/abstract, and 4) no information on any of the predefined outcome parameters.
Due to the exploratory nature of our study, we did not define specific hypotheses.This manuscript is based on the MOOSE Checklist of Meta-analyses and Observational Studies (see eTable 2 in the Supplementary Material). 21

Search terms for identification of studies
We searched the digital databases Embase, PubMed, CINAHL and PsycINFO, using a comprehensive search strategy consisting of a combination of subject headings and free-text words.The search strategy was developed together with a librarian (H.E.) experienced in systematic reviews.The final search strategy is provided in the Supplementary Material to ensure traceability and reproducibility of our results.To identify additional studies, we screened all references of eligible studies through the cited reference search of Web of Science and PubMed and applied the similar articles search of PubMed.The latest search was performed on January 22, 2021.

Study selection
Three investigators (H.C., A.V. and K.B.) screened the titles and abstracts of articles regarding inclusion and exclusion criteria.Two reviewers (H.C. and A.V.) independently assessed the full texts of all remaining studies and disagreements were resolved through discussion with a third reviewer (K.B.).

Risk of bias evaluation
We evaluated the risk of bias for every relevant outcome of all included studies using The Cochrane Collaboration's tool for assessing risk of bias. 22Two authors (K.B. and A.V.) independently assessed the risk of bias for all studies and resolved disagreements by discussion until consensus was found.A detailed description of the risk of bias assessment can be found in the Supplementary Material.

Analysis
We express dichotomous data risk ratios (RR) with 95% confidence intervals (CI).Data were pooled using a fixed-effects model.Heterogeneity (inconsistency) was identified through visual inspection of the forest plots.We used the I 2 statistic, which quantifies inconsistency across studies, to assess the consequences of heterogeneity on the meta-analysis.A considerable level of heterogeneity is indicated by an I 2 statistic of 50% or more. 235][26][27][28] Accordingly, a PAM or PAR score > 8, a GO-FAR score !14 and a PIHCA score 3% indicate that CPR would be medically futile.We also calculated the risk scores' specificity separately for each study.We applied narrative synthesis if data were not suitable for direct comparison.Statistical analyses were performed using the METAN package in Stata (Stata MP, version 15.1; StataCorp LP), and a two-sided p < .05 was considered statistically significant.

Identified studies
A total of 1966 records were identified through database searches and other sources.We removed duplicates (n = 86) and discarded 1621 studies after screening titles and abstracts.and in European countries such as Sweden, 4,28,31,49 and England. 40,41,44Study sample sizes ranged from 29 to 96,499 per trial.Half of the studies included hospitalized patients receiving CPR after IHCA 1,4,24,33,34,26- 28,41-43,46-49 and in three studies the study population consisted of hospitalized patients without IHCA. 25,30,40,52Yet, some studies included more specific patient populations such as elderly patients, 35,44 oncological patients, 36,38 critically ill patients, 39 severe burn patients, 32 or patients with multiple IHCA. 29One study only included patients with established DNR orders. 45

Definitions of futility
Twenty-seven studies included short descriptions or definitions of medical futility for CPR.These varied broadly in content and specificity, and rarely consisted of more than one or two sentences.Six studies defined futility as a very low likelihood of survival after CPR following cardiac arrest 1,10,33,39,45,52 with one of them specifying a "1% chance of surviving 2 months after CPR". 39Nine studies evaluating clinical risk scores solely presented a cut-off score indicating futility 24,42,44,52 or extremely low chance of survival with favorable neurologic outcome, defined as Cerebral Performance Category 1 10,47,48 or 1 to 2. 28,49 Ten studies provided unspecific definitions, defining futility either based on clinical conditions, e.g., age, metastatic cancer, or "acute or chronic impairments in almost any organ system in elderly patients" or based on an outcome, e.g., "prolonging the patient's suffering and therefore harming the patient". 25,30,31,35,36,38,40,45,50,51owever, none of these definitions included specific thresholds or criteria for futility.Four studies reported specific scenarios in which CPR would be futile, such as patients with a recurrent cardiac arrest or severe burn injuries. 29,32,34,37R code status in patients for whom CPR was deemed futile Four studies reported how many patients for whom CPR was deemed futile had a DNR code status.39,40,44,52 The definitions of futility among these studies and the rates of DNR code status varied considerably.In the study of Aarons et al. 40 junior doctors gave a statement concerning the appropriateness of resuscitation in case of IHCA for each included patient.Of all patients for whom CPR was perceived as futile, 27% (n = 24) had DNR orders.Stewart et al. 44 evaluated medical inpatients with a mean age of 84 years.CPR was judged futile if the patients' Pre-Arrest Morbidity (PAM) index scores were >4 and their Prognosis After Resuscitation (PAR) scores were >5 at the same time. Of these paients where CPR was considered to be futile, 44% (n = 17) had a DNR code status.Becker et al. 52 assessed 2889 patients hospitalized at the Division of Traumatology/Orthopedics or Internal Medicine.Futility regarding CPR was defined as a Good Outcome Following Attempted Resuscitation (GO-FAR) score ! 14and/ or a Clinical Frailty Scale (CFS) rating of !7.Of all patients where CPR was determined futile (n = 467), 69.2% had a DNR code status documented in their medical charts.Teno et al. 39 calculated a time-toevent prediction model in a sample of critically ill patients including diagnosis, age, number of hospital days before study entry, cancer diagnosis, neurologic function, and several physiologic measures all assessed on day 3 after study entry.CPR was determined as futile if the chance of 2-month survival was estimated to be 1% or less.The majority of those patients, i.e., 71% (n = 82) had a DNR order.

Meta-analysis of pre-arrest risk scores
The included studies examined different pre-arrest factors and prearrest risk scores based on these factors.Several risk scores were found in the systematic search to assess the pre-arrest risk of death during hospitalization after CPR for IHCA in individual patients.Eleven studies were included in the metaanalysis 4,24,41,43,[26][27][28][46][47][48][49] that assessed the following four prearrest risk scores: the Pre-arrest morbidity (PAM) index, the Prognosis After Resuscitation (PAR) score, the Good Outcome Following Attempted Resuscitation (GO-FAR) score and the Prediction of Outcome for In-Hospital Cardiac Arrest (PIHCA) score. In th supplementary material, we provide a detailed overview of these clinical risk scores (eTable 1).
In the studies included in the meta-analysis, the mean age varied between 60 years 48 and 72 years. 49Further, male gender ranged between 58 % 27,47 and 65%. 24,48All studies included inpatients receiving CPR after IHCA.Overall, the meta-analysis comprised 118,315 patients.
Five studies with 1621 patients reported PAM scores and inhospital mortality with a low risk of bias. 24,26,41,43,46The overall analysis showed that the PAM index was associated with a significantly higher risk of in-hospital death at a cut-off score of PAM > 8 (RR 4.10 [95 %CI 1.39-12.11]).Heterogeneity among trials was low (I 2 = 0.0%, p = .638).Specificity in the individual studies ranged from 98% to 100% (eTable 3).
Four studies with 1481 patients reported PAR scores and mortality with a low risk of bias. 26,41,43,46The PAR score was associated with a significantly higher risk of death until discharge at a cut-off score of PAR > 8 (RR 3.11 [95 %CI 1.59-6.05]).Heterogeneity among trials did not have a significant impact (I 2 = 54.5%,p = .086).Results of the PAM and PAR scores are shown in Fig. 2. Specificity in the individual studies ranged from 83% to 100% (eTable 3).
Five studies with 114,585 patients reported GO-FAR scores and poor neurologic outcome or in-hospital death with a low risk of bias. 4,27,28,47,48The GO-FAR score was associated with a significantly higher risk of poor neurologic outcome (CPC 1) and inhospital death at a cut-off score of GO-FAR ! 14 (RR 6.92 [95 % CI 6.42-7.46]).There was high heterogeneity among trials (I 2 = 81.1%,p < .001).Specificity in the individual studies ranged from 89% to 97% (eTable 3).
One study with 628 patients evaluated the PIHCA score 49 and poor neurologic outcome or in-hospital death with a low risk of bias.A very low or low ( 3% chance of favorable neurological survival) PIHCA score was associated with a significantly higher risk of poor neurologic outcome (CPC 2) and death until discharge (RR 11.46 [95% CI 1.65-79.61]).The specificity was 99% (eTable 3).Due to inclusion of only one study, heterogeneity could not be calculated.Results of the GO-FAR and PIHCA score are shown in Fig. 3.

Discussion
In this systematic review and meta-analysis, we investigated the concept, measures and application of medical futility regarding CPR after IHCA in clinical routine.We included 30 studies in the qualitative review and 11 in the meta-analysis.Aside from theoretical arti-cles from the field of medical ethics, we found only few clinical definitions of futility regarding CPR and no international consensus including specific definitions or criteria that allow for concrete implementation in clinical practice.Still, several studies proposed different pre-arrest objective risk scores for the definition of futility 44,52 or to assess extremely low chance of survival (with favorable neurologic outcome, i.e.8][49] In metaanalyses, these four risk scores, i.e.PAM index, PAR score, GO-FAR score and PIHCA score, were associated with in-hospital mortality and -in the case of the GO-FAR and PIHCA score -in-hospital mortality and poor neurologic outcome defined as CPC 1 and CPC 1 to 2, respectively.Several findings of this review need further discussion. First, we found a wide variation in definitions of futility, which were mostly either unspecific or limited to certain clinical conditions.In line with a systematic review on medical futility without focus regarding resuscitation 53 and current opinion of leading experts on the topic, 54 none of the studies in our review reported a definition including specific and well-defined criteria that would allow identification of patients  (continued on next page)   in whom the chance of futile CPR would be high in clinical practice.However, two studies used clinical risk scores and chose a specific cut-off to determine futility. 44,525][56][57] In public and academic discussions about futility, multiple fundamental issues have been raised that hinder an international consensus definition.These include ethical questions and concerns, societal values, cultural beliefs, legal challenges, and clinicians' responsibilities. 9,15,54,58o far, the most promising approach to evaluate quantitative futility was based on objective risk scores, i.e., the GO-FAR and PIHCA score regarding expected in-hospital mortality and neurological outcome.Such an approach to define futility quantitatively requires the definition of a specific threshold below which CPR would be assumed futile and a clear and clinically meaningful definition of "good outcome". 11Yet, determining such a cut-off value for use in clinical practice has ethical and clinical challenges and may depend on societal and patient factors and perspectives as well as preferences of patients and families. 55There may be differences in the perception what the best cut-off should be to define futility.When asked about their estimations of survival of patients in whom they perceived CPR to be futile, several physicians reported probabilities of over 5% and over 10%. 15,45Further, the two existing clinical risk scores additionally considering neurological outcome differ in their definition of good neurologic function, i.e.CPC 1 in the GO-FAR score vs. CPC 1 to 2 in the PIHCA score. 10,28Additionally, other meaningful clinical outcomes such as quality of life, self-reliance and severe health impairments, e.g., organ failures, need to be considered and incorporated into the concept of futility regarding CPR.
The concept of qualitative futility centers the patient's quality of life instead of quantitative parameters.This approach was only mentioned by two studies in our review. 45,51If applied consequently, this approach requires an evaluation of patients' subjective quality of life as well as their idea of a meaningful and fulfilling life, considering the potential adverse neurological consequences of CPR.Moreover, the latter seems to become even more prominent with advanced and invasive life-prolonging interventions, such as extracorporeal membrane oxygenation (ECMO), as these can be associated with neurological complications. 59 Only four studies 39,40,44,52 assessed the code status in patients in whom resuscitation was determined futile.The rates of DNR orders in these studies varied due to the high heterogeneity of patient samples, definitions and assessments of futility, making interpretation difficult.While Teno et al. 39 explicitly stated that surrogates and patients participated in decision-making, it is unclear if DNR orders in the other three studies 40,44,52 were unilaterally implemented by physicians or discussed with patients and/or surrogates.
[46][47][48][49] The PAM index and a modified version named the PAR score were developed four decades ago, about three decades after the invention of CPR. 60 Surprisingly, we only identified five studies evaluating these scores in relation to survival rates 24,26,41,43,46 and two further studies in relation to rates of DNR orders. 25,44Further, the studies used different cut-off scores for determining high risk of in-hospital death after CPR.In our metaanalysis, both the PAM and PAR score were associated with inhospital mortality, albeit with a lower predictive value than the newer GO-FAR score and its derivative, the PIHCA score.Most predictors of the PAM index are variables that were independently associated with mortality following CPR in a previous primary study 61 plus factors deemed relevant by the authors 24 and the PAR score is based on a meta-analysis of 14 studies. 62The GO-FAR score was developed based on multivariable analyses of a dataset of about 50,000 inpatients with IHCA. 1,27Further, the GO-FAR and PIHCA score predict survival with good neurologic outcome, which is defined as Cerebral Performance Category (CPC) 1, indicating good cerebral performance, in the GO-FAR score and as CPC 1 to 2 with 2 indicating moderate cerebral disability, in the PIHCA score.These aspects might have contributed to their better performance in our analysis.
Due to the PAM and PAR scores' above-mentioned limitations, we suggest focusing on the GO-FAR and PIHCA score in future research.Further studies are needed to validate the GO-FAR and/ or PIHCA score regarding their prognostic accuracy.As a next step, implementation of these scores in clinical practice is needed.Potential benefits and short-comings of using clinical risk scores for decision-making regarding code status need to be assessed, e.g., does decision-making change due to risk scores or is it coherent with clinical impression.Although clinical scores are never perfectly accurate in their prediction of outcome, they may reduce the influence of subjective factors that should not contribute to determining futility such as physicians' individual values and attitudes.Risk scores may help guide physicians in the difficult task of futility assessment to make this evaluation more objective, transparent, and hopefully reliable especially when physicians are still unexperienced.According to the online registration platform ClinicalTrials, there is currently one large randomized trial comparing code status discussions based on a checklist and risk assessment of futility using the GO-FAR score and the Clinical Frailty Scale with usual care (https://clinicaltrials.gov/ ct2/show/NCT03872154).

Conclusions
In summary, although in-hospital cardiac arrest occurs in about 2-3% of hospitalized patients and code status discussions and decisions are an integral part of clinical practice, there is only little research on consensus definitions of medical futility.While most clinicians would agree that there is a relevant proportion of patients in whom CPR is likely to be futile, our review found no established definitions of futility for use in clinical practice.International consensus regarding the definition of futility is lacking and tools for its assessment could improve objective code status discussions with patients.Communication about medical futility holds the potential of empowering patients to make informed decisions that are in alignment with their goals of care, avoiding unwanted physical and emotional suffering for them and their relatives, which may come along with unwanted life-sustaining measures and treatments in situations without realistic prospects for a desirable recovery in the individual case.A definition, criteria and measures suitable for the implementation of scoring systems to determine the likelihood of futility in specific clinical scenarios need to fulfill several requirements regarding acceptance, feasibility, and prognostic value, among others.Two recently developed clinical risk scores, the GO-FAR and PIHCA score, showed promising predictive values.However, further studies are needed to evaluate the implementation of such concepts and their assessments in clinical practice.

Fig. 1 -
Fig. 1 -PRISMA flow diagram of the study selection process.

Fig. 2 -
Fig. 2 -Forest plot showing the association of the PAM and PAR score and risk of in-hospital death.

Fig. 3 -
Fig. 3 -Forest plot showing the association of the GO-FAR and PIHCA score and risk of survival with impaired neurological outcome or in-hospital death.

Table 1
lists characteristics of the 11 studies included in the metaanalysis.Detailed information on the remaining 20 studies solely included in the qualitative synthesis are shown in eTable 2. Publication dates ranged from 1989 to 2019, and studies were conducted mostly in the United States,

Table 1 -
Summary of the studies included in the meta-analysis.

Table 1 (
continued) (continued on next page)

Table 1 (
continued) Abbreviations: CPR, cardiopulmonary resuscitation; IHCA, in-hospital cardiac arrest; OHCA, out-of-hospital cardiac arrest; PAM, Pre-Arrest Morbidity index; PAR, Prognosis After Resuscitation score; APACHE, Acute Physiology and Chronic Health Evaluation score; GO-FAR, Good Outcome Following Attempted Resuscitation score; PIHCA, Prediction of Outcome for In-Hospital Cardiac Arrest score; CPC, Cerebral Performance Category; DNR, do not resuscitate; AUROC, area under the receiver operating characteristic curve. The