Memory, attention and fluency deficits in COPD may be a specific form of cognitive impairment

Participants

We enrolled patients with moderate to severe COPD from a secondary care respiratory clinic in a medium-sized urban teaching hospital in the UK over an 8-month study period from June 2016 to April 2017. Moderate to severe COPD was defined as patients with a forced expiratory volume in 1 s (FEV₁) of <80% predicted.

As a comparator, patients with known cognitive impairment and a formal diagnosis of Alzheimer's dementia but without COPD were recruited from the Join Dementia Research Network. Control patients with a similar age profile and sex distribution but without a diagnosis of COPD or cognitive impairment were also recruited.

All subjects gave their informed, written consent in accordance with the Declaration of Helsinki and following approval from the National Health Service Local Research Ethics Committee (ref. 16/EM/0192)

Assessment of cognition

We used the Addenbrookes Cognitive Examination (ACE)-III to assess cognitive impairment. The ACE-III is a validated tool that can be used to establish areas of cognition affected in different types of dementias [6]. Cut-off scores for possible dementia and mild cognitive impairment are 82 and 88 out of 100, respectively. The ACE-III investigates six different domains of cognition: memory, attention, perception, visuospatial abilities, language and verbal fluency.

In a subgroup, data regarding baseline oxygen saturations, FEV₁ and FEV₁ % pred values were available. Data regarding FEV₁ and FEV₁ % pred were available for 29 participants (28 in the COPD group and one control). Data for baseline oxygen saturations were available for 28 participants (27 in the COPD group and one control). We completed further analysis comparing overall and domain-specific scores compared to baseline oxygen saturations, FEV₁ and FEV₁ % pred.

Statistical analysis

All statistical analyses were carried out in SPSS version 25 (IBM, Armonk, NY, USA). We analysed the difference in ACE-III total and domain-specific scores between groups. Dunn's pairwise tests were then carried out for two pairs of groups (COPD group compared to cognitive impairment group and COPD group compared to control group) for each analysis. The significance values were then adjusted using Bonferroni corrections, given multiple group comparisons.

ANOVA was then performed to control for cardiovascular disease (CVD) and smoking. A Dunnet's pairwise multicomparison test was then performed to compare the COPD and cognitive impairment group and the COPD and control group. We used linear regression with Pearson's correlation coefficient and Spearman rank correlation to compare the baseline saturations, FEV₁ and FEV₁ % pred data to overall and domain-specific ACE-III scores.

Results of narrative review

We retrieved 327 original articles of which 18 met the inclusion criteria. A further 94 were retrieved from the reference list of a prior systematic review of which 13 were included; 31 met the inclusion criteria for the final review. Table 2 describes summary characteristics of the included studies and major findings (see the supplementary material for detail and a Preferred Reporting Items for Systematic Reviews and Meta-Analyses diagram).

The included studies did not specify the type of dementia or major neurocognitive disorder present in subjects and rarely specified if dementing conditions with a known pathophysiology (such as alcoholic dementia) were excluded. The 31 included papers employed psychometric tests to establish the level of impairment in cognition and domains affected, but did not translate the results of these cognitive tests into clinical diagnoses. As different studies used different psychometric tests the definition of cognitive impairment was not standardised and differed in the different papers included in the review.

Four papers studied cognitive outcomes according to COPD severity [7–10]. However, one of these papers used “self-reported history of COPD” rather than pulmonary function testing to classify COPD severity [10]. Varying criteria were used to diagnose COPD in the included studies. These variations in definitions and methodological limitations affected the quality of included studies and the inferences that can be made, although it seemed cognitive impairment was increased in COPD patients and largely affected memory.

10 of the studies included in the review discussed effects of hypoxia or hypoxaemia on cognitive impairment, with two concluding that there is a relationship between arterial oxygen tension (P_aO2) and severity of cognitive impairment [7, 11]. Reduced P_aO2 particularly related to impairment in the areas of motor function, attention and processing speed. However, this association was inconsistent; Dodd et al. [1] found that cognitive function was only mildly impaired in patients without hypoxaemia. Antonelli-Incalzi et al. [12] included hypoxic–hypercapnic COPD cases and found that 48.5% had a form of cognitive deterioration; longer duration of hypercapnia and hypoxia was significantly correlated to cognitive impairment. Other mechanisms were considered by included studies; however, there was relatively little consistency to the findings (supplementary material). Since the review findings were somewhat limited, we felt validation work was necessary, which comprised the second part of our study.

Cross-sectional study

A total of 89 participants were recruited over the study period (44 with COPD, 17 with Alzheimer's and 28 controls). Recruitment was challenging, particularly for the known Alzheimer's dementia group, eventually resulting in use of a national network to enhance case ascertainment. Comparison of baseline characteristics demonstrated no significant difference between the three groups except for smoking exposure >10 pack-years and cardiovascular comorbidity, which were more frequently observed in the COPD group (table 3).

Patients with COPD had significantly lower total ACE-III scores compared to the healthy controls (p<0.001). 50% (95% CI 33.8–66.2%) of COPD patients demonstrated scores below the cut-off for cognitive impairment (88 out of 100). This compares to just 7% (95% CI 9.0–23.5%) of healthy control group participants. The distribution of total ACE-III scores is presented in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Distribution of Addenbrooke's Cognitive Examination (ACE)-III total scores. COPD: chronic obstructive pulmonary disease.

Domain-specific scores

In patients with COPD, memory and fluency subscores were significantly lower compared to healthy control group participants whilst other cognitive domains remained intact. Whilst patients with COPD had lower total ACE-III scores compared to healthy controls (COPD group median 87.5 (interquartile range (IQR) 84–91) versus control group median 96 (IQR 93–97)) scores were significantly higher than for patients in the known Alzheimer's dementia group (median 79, IQR 70–80).

There were significant differences in the overall ACE-III scores (p=0.019) and in the attention (p=0.004), memory (p=0.004) and fluency subscores (p<0.001) between the COPD group and the known Alzheimer's dementia group. Those with COPD scored significantly higher overall and in these domains.

Lower baseline saturations were significantly correlated with lower overall scores in the domain of fluency (p<0.01). Baseline saturations were not correlated with overall ACE-III score or with any of the other cognitive domains. Lower FEV₁ was correlated with lower ACE-III scores in the domains of memory (p=0.03) and fluency (p=0.04) but not overall ACE-III score (tables 4–6).

Prevalence of cognitive impairment

In people aged >65 years, dementias occur in approximately 5–8% and cognitive impairment in different degrees in 5–20% [14]. Included studies in the review showed a range of 2–52% of COPD patients to be affected, across age groups, although few studies used this as the specific outcome, choosing in many cases to focus on areas of cognition instead. In our cross-sectional study, the prevalence of cognitive impairment, as defined by the cut-off score in the ACE-III, in the COPD group was 50% (95% CI 33.8–66.3%). This prevalence rate is towards the higher end of those studies included in the review. There is wide variation in the prevalence of cognitive impairment in COPD in the studies included in our review. We hypothesise that the most important, likely reasons for this are the lack of consensus regarding the definition of cognitive impairment and use of different cognitive assessment tools; even studies that used the same cognitive tests had varying thresholds for the definition of cognitive impairment. Most commonly used was the Mini Mental State Examination (MMSE), which was used in 14 studies; it has been suggested that MMSE “should not be used as a diagnostic tool to identify dementia” [15] and may underestimate impairment compared to other cognitive tests [16, 17]. The MMSE score can be affected by age, education and cultural background [15]; studies did not always control for these factors when interpreting scores. This underestimation of impairment with the MMSE in many of the included studies in the review, coupled with the use of the ACE-III in our cross-sectional study, may have contributed to the prevalence of cognitive impairment we found in COPD patients.

Domains of cognition affected in COPD

Cognitive function is classified into domains: perception, attention, memory and learning, abstract thinking and executive function, language, and intelligence [18]. Of these, memory was the one most consistently impaired in both the review and our validation study [1, 12, 19–21], though some studies have reported broader deficits spanning memory, attention, symbolic representation, visual processing and shifting capacity [22]. Neuroimaging studies [21, 23] have found reduced perfusion of frontal and subcortical regions in COPD, possibly explaining executive dysfunction and attention deficits seen [1, 19]. Fluency was affected in our cross-sectional study; evidence from the review supported this.

From the included studies in the review, it was difficult to determine whether COPD associated with Alzheimer's or vascular dementia, perhaps through a confounding factor such as smoking, or was associated with a specific pattern of cognitive impairment unique to the condition. It is known that different areas of cognition are affected in different dementias; for example, in Alzheimer's dementia, memory, executive function, language, praxis and visuoconstructional abilities are primarily affected [24]. The consistent memory and fluency impairment of COPD seen in the review, and confirmed by our cross-sectional study, may be a specific form of cognitive impairment. Cognitive impairment, particularly in the memory and attention domains, may explain the poor adherence to inhaler therapy and health behaviour change seen in this population [7].

Dementia is a devastating condition in terms of quality of life. COPD patients are increasingly recognised as being at high risk for cognitive impairment and subsequent dementia. Clinically, this is important to allow for appropriate modifications in lifestyle and drug therapies. For example, if praxic skills are impaired in COPD, inhaled therapies would be difficult to self-administer; if verbal memory and attention are impaired, pulmonary rehabilitation may need modifications to bring greater benefit.

Risk factors for cognitive impairment related to COPD

Patients with COPD have progressive airflow limitation and emphysematous changes in the lungs, resulting in a ventilation–perfusion mismatch, potentially causing chronic hypoxaemia [25]. Alternatively, the association between cognitive impairment and COPD could be explained by altered, decreased cerebral perfusion mediated by vascular disease. There is widespread evidence suggesting hypoxaemia alone does not account for cognitive changes seen in COPD. Studies in both inpatients and outpatients have reported cognitive deterioration even in COPD subjects without severe hypoxaemia. In our cross-sectional study, lower baseline saturations were significantly correlated with lower ACE-III scores in the fluency domain but overall cognitive scores were not affected; as blood gases were not measured in our study, it is difficult to comment on the contribution of hypoxia as a mediating factor in development of cognitive impairment.

FEV₁ has been explored in relation to cognitive impairment in COPD. Cleutjens et al. [20] found worsening cross-sectional memory and numeric short-term memory were associated with worsening FEV₁. However, Frohnhofen et al. [26] found that only 78% of their cohort and 47% of those with cognitive impairment could adequately complete lung function testing. Therefore, it is difficult to accurately assess any association of cognitive impairment and FEV₁ if cognitively impaired patients cannot perform spirometry. In our cohort, lower FEV₁ was significantly correlated with lower scores on the ACE-III in the domains of memory and fluency. The mechanism for this, if unrelated to hypoxia, is not clear.

The relevance of comorbidity to cognitive impairment in COPD

Those with COPD are often smokers or ex-smokers and are older; they are likely to have significant comorbidities [22]. In keeping with this, in the cross-sectional study, baseline differences in comorbidities were observed between the COPD group and the control groups in CVD diagnosis and smoking history >10 pack-years. The effect of comorbidities in COPD and cognitive dysfunction was investigated by some of the included studies in the review. Dal Negro et al. [22] did not find any relationship between cardiovascular and metabolic comorbidity and cognitive dysfunction, concluding that comorbidities were unlikely to account for the cognitive dysfunction seen in COPD.

Villeneuve et al. [19] found that although patients with COPD had more vascular comorbidity, the number and extent of vascular comorbidities was similar between the COPD patients with and without cognitive impairment, implying that vascular comorbidities did not have a significant effect on cognition. In our cross-sectional study, adjustments for smoking and known vascular disease were made and differences in cognitive impairment remained. These adjustments were intended to act as a surrogate for undiagnosed cerebrovascular disease and, whilst imperfect compared to formal diagnosis by neuroimaging, imply that the cognitive impairment in COPD cannot be simply put down to being a cerebrovascular pathological process.

Strengths and limitations

The review augments prior work as our inclusion criteria allowed a wider overview of the research supporting interaction between COPD and cognitive impairment, highlighting that there are mediators of cognitive impairment that are specific to COPD, and determining key areas for future research. The strengths and limitations of our review are discussed in the supplementary material.

A major strength of the cross-sectional study is use of the ACE-III, which is sensitive and allows investigation of domain-specific cognitive function. Inclusion of known dementia patients was important, as this allowed us to show the pattern of cognitive impairment in COPD was different to that in Alzheimer's disease.

There are a number of limitations that should be considered when interpreting the findings of the study. This pilot study had a small sample size but despite this, shows interesting results suggesting further, larger scale studies into this area are warranted. Patients and controls were not matched on a case by case basis, in order to facilitate recruitment. Despite this, differences between groups were accounted for in the analysis. We did not employ biochemical, neuroimaging or neurophysiological methods to explore the underlying pathogenesis of cognitive impairment in patients with COPD. A larger study including people with other types of dementia (e.g. vascular dementia or Lewy body disease) could further clarify whether the cognitive domains affected for each type of dementia differ from the COPD group, thus giving further evidence supporting a COPD-associated pattern of impairment. Issues with recruitment, particularly in the established cognitively impaired population, precluded this in our centre.

We did not conduct blood gas analyses on patients involved in the study. This is a limitation as hypercapnia and hypoxia were found to be correlated with cognitive impairment in previous studies included in our narrative review. In further, larger studies into this area, it would be useful to conduct blood gas analyses at the time of data collection, and investigate whether there is any association between these and ACE-III scores.

Conclusions

The cross-sectional study demonstrates a high proportion of cognitive impairment in patients with COPD. This is in keeping with our findings in the literature review. The cognitive domains of attention, memory and fluency seem to be predominantly affected in our cohort. We cannot conclude that there is evidence of a specific, discrete type of cognitive impairment unique to COPD based on the results of this cross-sectional study and literature review alone; however, our findings further our understanding of the pattern of cognitive impairment in patients with COPD and may help explain the poor treatment adherence seen in this population. Further studies should seek to elucidate the underlying neurobiological mechanisms explaining this association.