Neurocognitive outcomes in adults following cerebral hypoxia: A systematic literature review

1.1Aetiology and prognosis

Hypoxia can develop as a result of anaesthetic accidents, carbon monoxide (CO) poisoning (Quinn, McGahee, Polittle et al., 2009), cardiac and respiratory arrest (Perez, Samudra & Aiyagari, 2016), pulmonary injuries and cerebral blood loss injuries (Schultz, Sepehry & Greer, 2018), following stroke (Ferdinand & Roffe, 2016), near drowning (Nucci, Lukasova, Sato & Amaro, 2016) and attempted suicide via hanging (Tazopoulou, Milijkovitche, Trucelle et al., 2016) as well as conditions such as COPD. Suffice to state, the pathophysiology and neuropathology is most clearly established for hypoxia following cardiac arrest, CO poisoning and stroke (Busl & Greer, 2010; Ferdinand & Roffe, 2016). In a recent study exploring the long-term prognosis of individuals following hypoxic injury, Heinz and Rollnik (2015) described the variability in outcome among survivors, acknowledging it can be either poor or good. Historically others have agreed, referring to the inconsistent pattern as representing a bimodal pattern, with a moderate skew in the direction of a poorer outcome (Bachman & Katz, 1997). Notwithstanding, significant periods of brain hypoperfusion are generally associated with poorer clinical outcomes (Busl & Greer, 2010).

In attempting to understand and predict the seemingly unpredictable pattern of outcome, some have proposed that structural, pathophysiology and cognitive changes following hypoxia are most dependent upon the nature and severity of the injury (Greer, 2006). In an extensive literature review published a decade ago, Anderson & Arciniegas (2010) identified the total duration of unconsciousness after hypoxic brain injury relates to long-term cognitive prognosis. Others including Caine and Watson (2000) strongly argued that impairment is related to the site and brain structures affected by the hypoxic event. Earlier studies however, reported no relationship between age, gender, cause, site of onset or severity of injury as predictive of long-term outcome (Levy, Caronna & Singer, 1985).

1.2Cognitive outcomes following hypoxic ischemic brain injury

A considerable amount of literature exists documenting the neurological, cognitive, behavioural and psychosocial outcomes of Hypoxic Ischaemic Brain Injury (HIBI). Among this patient group, parkinsonism, dystonia, seizures, decline in processing speed, memory, executive function and depression have all been reported (Allen, Tranel, Bruss & Damasio, 2006; Hopkins & Bigler, 2012; Porter, Hopkins, Weaver et al., 2002; Wilson, Staniforth, Till et al., 2014). The specific neurobehavioral and neurocognitive changes which arise following cerebral hypoxia vary from brain death to coma and minimally conscious states to varying degrees of cognitive impairment but nevertheless well described in the existing literature (Anderson & Arciniegas, 2010; Caine & Watson, 2000; Hopkins, Gale & Weaver, 2006; Lu-Emerson & Khot, 2010; Schultz, Sepehry & Greer, 2018; O’Connor, Verfaelli & Cermak, 1995; Wilson, 1996).

In an earlier seminal review (1966-1998) Caine and Watson (2000) observed memory deficits in over half (54%) of individuals. Changes in personality and behaviour were also described in 31.3% of cases. Sauve, Lewis, Blankenbiller, Rickabaugh, & Pressler (1996) employed a case-controlled design to examine cognitive impairment in chronic heart failure patients compared with matched healthy controls on a number of domains including orientation, attention, memory, executive function and speed of information and motor processing. Those with chronic heart failure scored poorer on all cognitive domains compared to age and gender matched healthy controls. Single case studies have also documented cognitive deterioration post injury. Parkin et al. (1987) observed on the effects of anoxic encephalopathy following cardiac arrest on a 43-year old female and describe memory loss, dysphasia and an overall decline in general cognitive function compared to premorbid ability. Similarly, in a case study reporting two adolescents following suicide attempts by hanging, Zabel, Slomine, Brady, & Christensen, (2005) document impairments in multiple cognitive domains including attention, expressive and receptive language, processing speed, visuospatial abilities, memory as well as executive functioning.

1.3Rationale for current review

With advances in acute medical treatment and improved resuscitation practice, survival rates are increasing among individuals who sustain even severe oxygen deprivation (Anderson & Arciniegas, 2010; Knot & Tirschwell, 2009). However, significant morbidity persists and increasingly individuals require longer-term support beyond acute care. Nevertheless, the potential physical, cognitive and psychological consequences of cerebral hypoxia remain variable. To date there has been no published systematic review documenting the cognitive outcomes of adult survivors following Hypoxic Ischaemic Brain Injury (HIBI). Four extensive albeit non-systematic reviews have been published (Anderson & Arciniegas, 2010; Caine & Watson, 2000; Ferdinand & Roffe, 2016; Schultz, Sepehry & Greer, 2018). In addition, one unpublished thesis study which although systematic had identifiable gaps in terms of search terminology and subsequent output (Crooks, 2014). Accordingly, systematic data synthesis could provide an enhanced understanding of the impairment patterns observed among individuals with varying aetiology and course as well as facilitated improved prognostics in subacute and post-acute/rehabilitation settings. This paper aims to systematically review the peer reviewed published literature on cerebral hypoxia related to observed neurocognitive outcomes in adults.

2.1Search strategy

This systematic review was prospectively registered with Prospero (2017: CRD42017054692) in keeping with PRISMA-P 2015 statement (Moher, Shamseer, Clarke et al., 2015). Relevant literature was sourced via a systematic search of six common databases namely Medline, PsycINFO, Web of Science, CINAHL, Embase and Cochrane. Studies were identified using MESH headings. Limits placed on the search specified time (January 1970-December 2016), language (English), population (humans) and age (>18 years). Dissertation and conference proceedings were excluded, and duplicates removed. The following search terms were employed: Ischemi* OR ischemic injury OR hypoxic ischemic OR hypoxi* OR anoxi* OR cerebral hypoxia OR brain hypoxia OR anoxic brain injury OR anoxic encephalopathy OR hypoxic brain damage OR hypoxic brain injury OR hypoxic encephalopathy OR cerebral anoxia OR brain anoxia AND Cogniti* OR cognitive function OR cognitive disorders OR neuropsychological impairment OR cognitive impairment OR mild cognitive impairment OR mild neurocognitive disorder OR cognitive decline OR mental deterioration OR attention OR memory OR perception AND sequela* OR outcome* OR survi* OR rehabilitation OR neurorehabilitation NOT sequela* OR outcome* OR survi* OR rehabilitation OR neurorehabilitation.

The search involved three sub-searches (see Fig. 1). A total of 2,962 articles were identified through database searching comprising 759 articles from Medline, 155 from PsycInfo, 548 from Web of Science, 330 from CINAHL and 1170 from Embase. A primary screen of article titles for relevance based on Eligibility Criteria 1, resulted in the exclusion of 2,640 articles due to their use of experimental design (medication or therapy intervention), focus on neurochemical status or neuroimaging. Of the remaining 322 articles, duplicate studies were removed (n = 145) leaving 177 articles reviewed by abstract. Using Eligibility Criteria 2, of the 127 abstracts reviewed, 41 were excluded on the basis that no standardised neuropsychological tests were used to measure cognitive outcomes, 19 were experimental design with very limited comparative data, 6 used a non-English speaking population, 28 were review articles, 17 did not specify hypoxia as the primary diagnosis, 5 were animal studies, 3 were editorial correspondences, 3 involved individuals under 18 years, 1 article was published outside of the specified timeframe and 1 was only available as an abstract.

Fig. 1

Systematic Search Strategy.

2.2Eligibility Criteria 1: Criteria applied at title review

Inclusion Criteria: (1) Studies involving human, adult populations; (2) Papers published in peer reviewed journals from 1st January 1970; (3) Studies which examined neurocognitive outcomes using validated, objective and performance-based measures.

Exclusion Criteria: (1) Papers which were review articles, conference proceedings and letters to the editor; (2) Studies involving non-human populations; (3) Individuals aged <18 years.

The full text of the remaining 54 articles underwent screening by two independent reviewers based on Eligibility Criteria 2 outlined below.

2.3Eligibility Criteria 2: Applied at full text and abstract review

Inclusion Criteria: (1) Studies involving human, adult populations; (2) Papers published in peer reviewed journals since 1970; (3) Studies which examined neurocognitive outcomes using validated, objective and performance based measures.

Exclusion Criteria: (1) Studies involving populations with neurodegenerative condition including sleep apnoea and hypoxic injury as a result of high altitude climbing; (2) Treatment or experimental studies; (3) Studies using self-report measures or rating scales of cognitive function; (4) Studies which used translated measures originally normed on a UK or American population; (5) Studies involving non-English speaking participants; (6) Studies which administered neuropsychological tests which cannot be compared; (7) Studies reporting data which cannot be compared with other available normative data.

Of the 54 articles, 13 were excluded as they involved a non-English speaking population, 10 did not use objective measures of cognitive function, 1 used neuropsychological tests which were not comparable to any other studies, 1 was a descriptive paper only and 1 was a comments paper from a conference. The measured weighted Cohen’s Kappa for the two independent raters was high (k = 0.93). There was a discrepancy between reviewers on the relevance of two papers (Browne, Halligan, Wade & Taggart, 2003; Fitzgerald, Aditya, Prior, McNeill & Pentland, 2010) due to availability of data in the articles however this was resolved by a third independent reviewer.

Two independent reviewers agreed that 28 articles would be included in the final synthesis based on the eligibility criteria outlined for full article review. It was not possible however, to extract the data from ten papers due to the method used to report the outcomes of cognitive testing (composite scores) (Alexander, Lafleche, Schnyer et al., 2011; Allen, Tranel, Bruss & Damasio, 2006; Grubb, Fox, Smith et al., 2000; Fitzgerald, Aditya, Prior et al., 2010; Lim, Alexander & LaFleche, 2004; Lim, Verfaellie, Schnyer et al., 2014; Myers, Hopkins, Deluca et al., 2008; O’Reilly, Grubb & O’Carroll, 2003; Porter et al., 2002; Sauve, Lewis, Blankenbiller et al., 2009; Speach, Wong, Cattarin et al., 1998). Where possible the raw data was requested from the authors, but this was either not available or the data that was provided was not amenable to comparison with the data extracted from the remaining articles. The primary data was not documented in one paper (Browne et al., 2003), however it was available from an earlier paper (Taggart, Browne, Halligan & Wade, 1999) and therefore included in data extraction. Accordingly, a total of eighteen studies (9 case studies; 9 group studies) were included in the final synthesis.

2.4Data extraction

Where studies reported standardised scores, it was possible to extract data for analysis. For case studies the scores were converted to z-scores based on the population mean and standard deviation as reported for each test. Comparability of group studies was facilitated by converting reported mean performance to Cohen’s d representing the overall effect size. Prior to analysis and in order to interpret results, scores of neuropsychological tests which fell equal to or more than 2 S.D below the mean of the normative population were defined as having significant impairment in that domain, while impaired functioning was indicated for performance equal to or greater than 1 S.D. For those case studies which employed a longitudinal design, the latest available time point was included in data analysis. Where relevant, results were considered significant at the 0.01 level.

3.1Quality appraisal

The Newcastle-Ottowa Scale adapted for Cross-Sectional Studies (Wells, Shea, O’Connell, Peterson, Welch, Losos & Tugwell, 2008) was employed as a framework to assess the overall quality of articles included in the final synthesis. A second reviewer also independently assessed the quality of articles and the measured weighted Cohen’s Kappa for the raters was high (k = 0.90). When the quality criteria were applied to each study, the most frequently fulfilled quality requirement of the Newcastle-Ottowa Scale related to the method of assessment outcome employed during the study and the statistical analysis used (see Table 1).

3.2Recruitment and sample information

A summary of the eighteen papers included in the final review is provided in Table 1. For case studies (n = 9), age at testing ranged from 18–67 years with a mean age of 37.66 (S.D. = 13.59). The average age range of those who participated in group studies (n = 9) was 30.22–70 years with a mean age of 48.82 (S.D. = 13.29). Time interval from the hypoxic event to cognitive assessment varied among the studies ranging from 1 month to 28 months. Aetiology differed among the studies and included heart failure, anterior communicating artery (ACoA), acute respiratory distress syndrome, respiratory arrest, carbon monoxide poisoning, in hospital cardiac arrest, attempted hanging, coronary artery bypass surgery, near downing, status epilepticus, anaesthetic accident, hantavirus pulmonary syndrome, anaphylactic reaction and chlorine gas poisoning.

All eighteen studies used standardised neuropsychological assessment to examine cognitive outcomes. Specifically, 14 used the Wechsler Adult Intelligence Scale Revised (WAIS-R) (Wechsler, 1987), 10 used the Rey Osterrieth Complex Figure Test (ROCFT) (Rey & Osterrieth, 1993), 9 used Trail Making Test (TMT) version A and B (Reitan & Wolfson, 1985; Reitan & Wolfson, 1995), 8 used the Rey Auditory Verbal Learning Test (RAVLT) (Schmidt, 1996) and 7 administered the Wechsler Memory Scale –Revised (WMS-R) (Wechsler, 1987). The remainder of the neuropsychological measures used varied and included The Boston Naming Test (BNT) (Kaplan, Goodglass & Weintrab, 1983), The Wisconsin Card Sorting Test (WCST) (Heaton, Chelune, Talley et al., 1981), The Rivermead Behavioural Memory Test (RBMT) (Wilson, Cockburn & Baddeley, 1985) and the WAIS-R Digit Span subtest (Wechsler, 1987).

3.3Overall intellectual functioning

Intellectual functioning was assessed using the WAIS-R in 19 cases and 5 group studies (n = 166). In the main, studies reported relatively preserved function with some exceptions of marked impairments. The pooled mean FSIQ for the group studies was 98.7 (S.D. = 1.15) and comparable to normative data (d = 0.12). The mean FSIQ of the case studies was 83.95 (S.D. = 15.50) indicating impairment relative to the general population with a large effect size calculated (d = 1.05). Analysis using a One Sample t Test showed a significant difference in mean FSIQ [t (18) = –4.15, p = 0.001], VIQ [t (18) = –3.89, p = 0.001] or PIQ [t (18) = –2.736, p = 0.014] of the reviewed studies. It is probable some of the reported decline in cognitive function observed in the case studies may be accounted for by lack of use of standardised measures of premorbid ability (Medalia, Merriam & Ehrenreich, 1991; Wilson, Harpur, Watson & Morrow, 2003). Notwithstanding, the observed decline in cognitive function within the remaining studies was not in keeping with premorbid estimates (Auerbach & Hodnett, 1990; Hopkins, Gale, Johnston et al., 1995; Armengol, 2000).

3.4Assessment of language

Language ability was measured using the Boston Naming Test in two case studies (n = 4) (Armengol, 2000; Baggett, Kelly & Korenman, 2003) and one group study (n = 93) (Moser, Cohen & Clark, 1999). In the case studies, all but one individual showed severely impaired ability. The overall mean for the case studies was 36.25 (S.D. = 20.0) with a large effect size (d = 1.27) well below the normative mean. For the group studies, a small effect size was observed when performance was compared to the study’s control group (matched on age and gender) (d = 0.29) relative to the normative sample (d = 0.37). Despite the variability between the case and group studies, a One Sample t Test showed no significant difference in overall ability ([t (3) = –1.60, p = 0.208]).

3.5Memory and learning

On formal neuropsychological measures of memory function, the results showed varied cognitive sequelae with predominant disturbance in verbal ability observed across all studies. Using the RAVLT, four case studies (n = 12) demonstrated rapid loss of verbal memory and overall performance below the normative population on a learning task (M = 4.75; S.D. = 1.76) and delayed recall (M = 1.44; S.D. = 2.19). Comparatively, the mean performance on the learning trial of the RAVLT among the five group studies (N = 202), was below the normative mean (M = 5.7; S.D. = 0.33) and reflected a medium-large effect size (d = 0.74). For studies which administered the recognition trial of the RAVLT all individuals performed better on recognition than delayed free recall (Hopkins, Gale, Johnston et al., 1995; Hopkins, Larson-Lohr, Weaver et al., 1998; Hopkins, Weaver, Pope et al., 1999). Four cases also assessed verbal memory (n = 8) using the WMS-R and had a mean performance of 84.5 (S.D. = 25.0) reflecting ability below the normed mean with a medium effect size (d = 0.74) observed. For three group studies (n = 141) pooled mean performance on verbal memory tasks of the WMS-R was 90.5 (S.D. = 5.84) demonstrating ability below the normative population with a large effect size calculated (d = 0.82).

Visual memory ability as assessed by the Wechsler Memory Scale Revised (WMS-R) was comparable to normative data among the case studies (d = 0.21) and group studies (d = 0.01). However, impaired performance was observed among group studies on the Rey Osterrieth Complex Figure Test (ROCFT). Two group studies administered the ROCFT immediate visual memory task (n = 121) with a mean group performance of 15.2 (S.D. = 1.15), reflecting impairment below normed data with a large effect size observed (d = 0.95). On delay, impairment persisted with a mean of 13.4 (S.D. = 3.4) below the general population and reflecting a medium effect size (d = 0.67).

Among studies which employed the Rivermead Behaviour Memory Test (RBMT), impaired performance was observed in both case and group studies. Both case studies (d = 3.02) and group studies (d = 2.23) yielded a large effect size, with performance below the test norms. Finally, as Table 3 illustrates, the One Sample t Test showed no statistically significant difference between the mean performance across case studies and the group studies on all tasks of memory, with exception of RAVLT delayed verbal recall.

3.6Attention and concentration

Formal assessment of attention was limited across studies and employed the WAIS-R digit span subtest. Three case studies (n = 4) had an overall mean performance of 7.5 (S.D. = 2.9), with a medium effect size (d = 0.78) below the normative mean. The pooled mean performance for three group studies (n = 97) using the digit span was 8.8 (S.D. = 0.56), reflecting a medium effect size (d = 0.45) below the mean performance of the normative sample. There was no significant difference in the mean performance of the case studies and group studies ([t (4) = –0.921, p = 0.425]).

3.7Visuospatial construction abilities

The copy subtest of the ROCFT was used to assess visuo-spatial and constructional abilities in three cases (n = 6) and five group studies (n = 166). The mean performance of the case studies was 28.5 (S.D. = 9.3) with a medium effect size calculated (d = 0.62) and reflected performance below normative data. The mean performance of the group studies was 30.8 (S.D. = 2.09), with a small effect size calculated compared to the normative population. In the main, overall performance reflected ability in the normal range, although there were some individual exceptions which showed impairment. Notwithstanding, One Sample t Test analysis showed no significant difference in mean performance between the case and group studies ([t (5) = –0.606, p = 0.571]).

3.8Executive function

Formal assessment of executive function using the TMT, showed ability below the normative mean on Trails A, a simple task of visual search and sequencing and impaired performance on Trails B, a task of attentional switching. Three case studies among five individuals had a mean performance on TMT A of 39.4 (S.D. = 9.3) with large effect size calculated (d = 1.12) illustrating performance below the normative sample. For six group studies (n = 265) the pooled mean performance on TMT A was (M = 34.19; S.D. = 4.44) with a medium effect size calculated (d = 0.31) again indicating performance below the normative population. On TMT B the case studies (d = 1.93) and group studies (d = 1.61) showed a large effect size, with ability below the normative mean. One Sample t Test revealed no significant difference in performance between the case studies and group studies on either the TMT A ([t (6) = –3.104, p = 0.021]) or TMT B [t (6) = –2.904, p = 0.027]).

Executive function was also assessed using the original Wisconsin Card Sorting Test (WCST) among three case studies (n = 6) and 2 group studies (n = 35). The pooled mean number of perseverative errors among the case studies was 32.25 (S.D. = 23.4) with a large effect size (d = 1.39) compared to normative data. For group studies the mean number of perseverative errors was 10.46 (S.D. = 10.29), and comparable to the normative population (d = 0.12). A One Sample t Test revealed no significant difference between the mean performance of case studies and group studies ([t (3) = 1.86, p = 1.60]).

4.1Review limitations

While this review provides a systematic synthesis of the neurocognitive outcomes in adults following cerebral hypoxia, the reliability and generalisability of the results should be carefully considered in light of a number of notable limitations. Firstly, the quality assessment performed using the Newcastle Ottawa Quality Assessment Scale noted that included articles employed standardised age normed neuropsychological assessments. Nonetheless, review findings were derived from small sample case or group studies. Thus, it is possible that included studies may not be representative of the larger HIBI clinical population particularly those presenting following severe cerebral anoxia. In addition, few studies utilised any measure of estimated premorbid cognitive ability. Furthermore, studies often did not offer details on the method of recruitment and selection and where studies had non-respondents, no details were provided to allow for comparability.

Heterogeneity in relation to age, aetiology and severity of the hypoxia and the time interval between hypoxic incident and neuropsychological assessment, further restrict the comparisons which could be made among the studies. Of note, none of the included studies utilised measures of performance validity which would now be commonly employed in routine clinical practice (British Psychological Society practice guidelines, 2009; Bush, Ruff, Tröster et al., 2005). In addition, in terms of age, there was a significant difference between included group and case studies. Previous research has considered age a contributor to the degree of impairment following cerebral hypoxia (Hopkins, Gale, Johnston et al., 1995) and recommended younger individuals (<30 years) are most appropriate for rehabilitative interventions post hypoxia (Grosswasser, Cohen & Costeff, 1989). However, age related effects have not been reported unequivocally (Levy et al., 1985; Greer, 2006). With varied aetiology of injury, previous reviews identified that the extent of cognitive impairment following hypoxia is related to the cause of the injury (Anderson & Arciniegas, 2010; Busl & Greer, 2010). However, others have argued that severity and aetiology of hypoxia is not associated with long-term neuropsychological outcomes (Hopkins et al., 1995). Hopkins et al. (1995) noted remarkably different outcomes among the three individuals described in a case study and yet reported comparable patterns and severity of initial brain insult. Armengol (2000) reported the length of coma was not predictive of impairments in neuropsychological or psychosocial outcomes and Gale et al (1999) proposed anatomical changes post injury visualised by imaging were related to degree of impairment rather than aetiology of insult. Comparatively, by exploring the outcomes of those with a traumatic brain injury (TBI) and those with an anoxic brain injury, Hopkins, Tate & Bigler (2005) observed no significant difference in overall cognitive sequelae, which it could be argued highlights the limited influence of aetiology on neuropsychological outcomes.

Interval from injury to testing also varied considerably among the studies and it is possible those tested in an earlier, more acute phase could show greater impairment on testing, which may not be reflective of their neuropsychological function overtime. However, previous studies have estimated that greatest recovery takes place in the initial three months post hypoxic injury, which may suggest the differences in interval time among the studies in the present review had limited influence (Allen, Tranel, Bruss & Damasio, 2006; Lim, Alexander & LaFleche, 2004).

Findings are also constrained by the varied use of neuropsychological tests across the studies. While some used extensive neuropsychological batteries, others employed briefer screening assessments which may be less sensitive at detecting impairment of higher cognitive functions (Porter, Hopkins, Weaver et al., 2002). It is also possible that inaccessible data could have contributed further to clarifying the overall picture of neuropsychological outcome following hypoxia. Finally, while the method of calculating effect size offered a means of comparing to normative data, some have cautioned that the use of qualitative categories as described by Cohen (1981) can result in the underestimation of impact of what Cohen considered a small effect size (Glass, McGaw & Smith, 1981).