MEDICINE
ORIGINAL ARTICLE
Cognitive Reserve and the Risk of Postoperative Cognitive Dysfunction A Systematic Review and Meta-Analysis Insa Feinkohl, Georg Winterer, Claudia D. Spies, Tobias Pischon
SUMMARY Background: Post-operative cognitive dysfunction (POCD) occurs in 10 to 54% of older patients during the first few weeks after surgery, but little is known about risk factors predisposing to POCD. Methods: Systematic literature review and meta-analysis of cognitive reserve indicators and POCD risk. Results: Fifteen studies on 5104 patients were included. Follow-up periods spanned 1 day to 6 months. Educational level was the most commonly assessed cognitive reserve indicator, and a longer time spent in education was associated with a reduced risk of POCD (relative risk [RR] per year increment 0.90; 95% confidence interval: [0.87; 0.94]), i.e. each year increase in education was associated with a 10% reduced risk. Similar findings were made for some analyses on education as a categorical predictor (high school versus further/higher education, RR 1.71, [1.30; 2.25]; lower than high school versus further/higher education, RR 1.69, [1.17; 2.44]) though risk was equivalent for patients with high school education and those with lower than high school education (RR 1.02; [0.78; 1.32]). Conclusion: Patients with a relatively higher level of education are at reduced risk of POCD. Risk stratification of surgical patients according to educational level may prove useful. ►Cite this as: Feinkohl I, Winterer G, Spies CD, Pischon T: Cognitive reserve and the risk of postoperative cognitive dysfunction—a systematic review and meta-analysis. Dtsch Arztebl Int 2017; 114: 110–7. DOI: 10.3238/arztebl.2017.0110
Molecular Epidemiology Research Group, Max Delbrück Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin-Buch: Insa Feinkohl, PhD, Prof. Dr. med. Pischon, MPH Charité – Universitätsmedizin Berlin: Prof. Dr. med. Winterer, Prof. Dr. med. Spies, Prof. Dr. med. Pischon, MPH MDC/BIH Biobank, Max Delbrück Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin-Buch and Berlin Institute of Health (BIH), Berlin: Prof. Dr. med. Pischon, MPH
110
ost-operative cognitive dysfunction (POCD) occurs relatively frequently, in 10 to 54% of patients during the first few weeks after surgery (1). It is usually transient (2), but unlike for post-operative delirium (POD), clear diagnostic criteria are lacking for POCD (3, 4). Despite its high prevalence, POCD is underresearched and well-established risk factors for POCD are few and far between (for a review, see [2]) so that at present the cognitive risk of a surgical patient is unpredictable. Recent research has shown that diabetes (5) and pre-existing cognitive impairment (6) may predispose patients to POCD. Compared with these types of clinical risk factors, however, research into the contribution of cognitive reserve to POCD has essentially been neglected. Cognitive reserve is a theoretical construct that aims to explain links between factors such as a lower level of education, lower socioeconomic status (SES), or lower pre-morbid cognitive ability and an increased risk of cognitive impairment in older age (7–11). The account assumes that people differ in their ability to functionally ‘buffer’ neuropathological insult due to aging and disease according to their cognitive reserve capacity (12–14). Simply put, brain networks of high-reserve individuals are thought to be better able to cope with disruptions due to working more efficiently and more flexibly compared with low-reserve individuals. Neuropathological burden may further be compensated for through recruitment of novel brain networks (13). POCD is known to negatively impact on subjective cognitive function and quality of life in affected patients (15, 16). Studies suggest that it also increases the risk of dementia and mortality (17–19). POCD is thus a cause for concern from a public health perspective that exceeds problems associated with cognitive deficits alone. With a lower cognitive reserve capacity as a predictor of age-related cognitive impairment, it appears reasonable to expect an association with POCD. A lower level of education is indeed frequently discussed as a contributing factor to POCD, though empirical evidence is rarely mentioned (2, 20–23). If such evidence was to be confirmed, measures of cognitive reserve could supplement cognitive risk prediction on the basis of clinical risk factors. Because low cognitive reserve
P
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
MEDICINE
could reasonably constitute the starting point of a causal chain leading up to POCD, the identification of cognitive reserve parameters as risk factors for POCD would further add to our understanding of the processes underlying the condition. Here, we aim to integrate the current epidemiological evidence on cognitive reserve and the risk of POCD in view to providing guidance for clinical practice.
Methods Systematic search strategy An electronic search (eTable 1, eBox 1) was performed by one investigator (IF) in accordance with the MOOSE and PRISMA guidelines (24, 25). Study selection Studies were eligible for inclusion if they ● followed a prospective study design, ● included human adults undergoing surgery (age ≥18 years), ● had full texts published in English ● reported original data on associations of cognitive reserve indicators (eTable 1) with POCD in the form of odds ratios or relative risks (RR; both termed RR in the present analysis) or as descriptive data that allowed calculation of RR. Any operationalization of POCD qualified for inclusion provided it was based on performance-based neuropsychological assessment. Data extraction Fully adjusted RR statistics were extracted unless no adjustment was made. If more than one article reported on the same sample, the article with the most complete reporting was selected. Data were extracted on the longest follow-up period. For 2 studies with multivariate-adjusted data at 7 day follow-up but not at 3 months, the 7 day follow-up was selected (26, 27). For one study comparing three levels of cognitive change, ‘severe deterioration’ was used to represent POCD (28). Enquiries were made to corresponding authors for unreported information. Data synthesis and analysis Studies were analyzed separately for each cognitive reserve indicator. We used the standard I2 index to identify statistical heterogeneity (29) and inverse variance fixed-effect models to calculate summary estimates of RR (95% CI) in meta-analyses across studies. Forest plots were generated to present pooled estimates. The main meta-analyses were repeated using random-effects models (eBox 2). Potential sources of heterogeneity were explored in subgroup and metaregression analyses. Review Manager 5.3 and SAS Enterprise Guide 4.3 were used.
Results The search retrieved 109 unique articles (eFigure 1); an independent search identified a further 28 articles. Overall, 64 full text articles were assessed. Of these, Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
40 did not meet our inclusion criteria and 9 articles (30–38) were excluded due to suspected duplicate reporting (19, 26, 27, 39). In total, 15 articles were included. The included studies originated in Europe, USA, Australia, and Asia (eTable 2). Surgical procedures included cardiac surgery (n = 7) and non-cardiac surgery (n = 8) under general (n = 9) or a combination of general and regional anesthesia (n = 4) (where reported) (Table). A total of 5104 patients were analyzed. Sample characteristics varied substantially between studies. The proportion of males ranged from 26% to 79% and the mean sample age from 51 to 75 years (mean 63 ± 8 years). Patients were followed up for between 1 and 180 days after surgery (median: 25 days; interquartile range: 7 to 45 days). All studies except one (40) applied detailed batteries of neuropsychological tests, though the criteria used to define POCD were heterogeneous. POCD occurred in 8% to 67% of patients. Findings of included studies and meta-analysis a) Years of education—Eight articles reported data on years of education. The mean years of education in these studies ranged from 8 years in 2 studies from Italy and China (28, 40) to 14 years in 2 US studies (17, e1) (mean 12 ± 3 years). When effects were pooled, each year increase in education was associated with a 0.90 risk of POCD (RR 0.90 per year increment; 95% CI: [0.87; 0.94]; p16 years of education” was equated with “further/higher education.” b) High school education versus further/higher education—When effects were pooled across 4 studies (19, 27, e2, e3), high school level of education was associated with a 71% increased risk of POCD compared with a higher level of education (RR 1.71 [1.30, 2.25]; p 70
56 ± 11
6 days
1 day
7 days
3 to 30 days
3 months
RR: 2.00 [1.33; 2.98]
RR: 0.83 [0.74; 0.93]
4. RR: 2.44 [0.92; 6.50]
3. RR: 2.51 [0.90; 7.03]
1. RR: 1.47 [0.46; 4.69] 2. RR: 1.03 [0.59; 1.78]
RR: 1.01 [0.96; 1.07]
RR: 0.93 [0.85; 1.00]
Exposure association with POCD*1
1. Middle school educationversus lower than 1. RR: 1.07 [0.78; 1.32] middle school education 2. Middle school education versus further/higher 2. RR: 1.80 [0.12; 27.25] education 3. RR: 1.61 [0.11; 22.70] 3. Lower than middle school education versus further/higher education
≤16 versus >16 years of education
Years of education (mean 8 ± 3)
1. Illiterate versus literate 2. High school education versus lower than high school education 3. High school education versus further/higher education 4. Lower than high school education versus further/higher education
NART score (mean 116 ± 10)
Years of education (mean 13 ± 4)*2
69 ± 8*2 6 weeks
Exposure variable
Mean age ± standard deviation Follow-up /median (interquartile range) (years)
Data are on analysis sample unless otherwise indicated. CABG, coronary artery bypass grafting; CI, confidence interval; CR, cognitive reserve; NART, National Adult Reading Test; POCD, postoperative cognitive dysfunction; RR, relative risk; SD, standard deviation; *1definition of POCD varied between studies; see eTable 2 in Supplementary Material; *2based on total sample enrolled in the study (data on analysis sample unreported) .
316
N
Mathew et al. 2013 (e7)
Author, year (reference)
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
MEDICINE
FIGURE a) Years of education Study (reference)
log (Relative risk)
SE
Weight
IV Fixed effects [95% CI]
Year
–0.6382 –0.1009 –0.09 –0.1381 –0.1685 –0.0161 –0.0774 –0.1863
0.2736 0.0413 0.0816 0.0674 0.0514 0.0416 0.0425 0.0586
0.5% 21.2% 5.4% 7.9% 13.7% 20.9% 20.0% 10.5%
0.53 [0.31, 0.90] 0.90 [0.83, 0.98] 0.91 [0.78, 1.07] 0.87 [0.76, 0.99] 0.84 [0.76, 0.93] 0.98 [0.91, 1.07] 0.93 [0.85, 1.01] 0.83 [0.74, 0.93]
2001 2007 2007 2008 2008 2010 2013 2014
Total [95% CI] 100.0% Heterogenity: Chi2 = 12.49. df = 7 (p = 0.09), I2 = 44% Test for overall effect: Z = 5.35 (p 16 years of education
239/440 in no-POCD group had ≤16 years of education
RR 0.83 (0.74; 0.93)
Exposure association with POCD
17/22
15/22
STROBE score*3
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
X
XI
80
Ni et al. (e3), 2015
78
N followed up
46%
% male
Intrathecal anesthesia
Total knee arthroplasty
Type of surgery and anesthesia
70 ± 4
Mean age ± SD/median (interquartile range) (years)
6 days
Follow-up
5 neuropsychological tests; MMSE
Cognitive measurement
POCD in n = 15/585 (24.8%)
Control group n = 20
POCD defined as ≥2 RCI*2 on >2 tests
Definition and incidence of POCD
1. Middle school education (n = 19/78)versus lower than middle school education (n = 56/78) 2. Middle school education versus further/ higher education (n = 3/78) 3. Lower than middle school education versus further/higher education
Exposure variable
None
Covariates
Descriptive data
Original statistical reporting
1. RR 1.07 (0.78; 1.32) 2. RR 1.80 (0.12; 27.25) 3. RR 1.61 (0.11; 22.70)
3/63 no-POCD patients with further/higher education
15/63 no-POCD patients with middle school education
45/63 no-POCD patients with lower than middle school education
0/15 POCD patients with further/higher education
4/15 POCD patients with middle school education
11/15 POCD patients with lower than middle school education
Exposure association with POCD
19/22
STROBE score*3
Data are on analysis sample unless otherwise indicated. APOEe4, apolipoprotein e4 allele; ASA class, American Association of Anesthesiology class of physical status; CABG, coronary artery bypass grafting; CR, cognitive reserve; MI, myocardial infarction; MMSE, Mini Mental Status Examination; NART, National Adult Reading Test; NYHA, New York Heart Association Functional Classification for cardiac risk; OR, odds ratio; RR, relative risk; SD, standard deviation; TMT-A, Trail-Making Test-A. *1 based on total sample enrolled in the study (data on analysis sample unreported) *2 RCI, Reliable Change Index (sometimes referred to as ‘z-scores’ in original publications). Formula to obtain RCI for each patient: (change score of patient – mean control group change score) / SD of control group change score (see [4]). *3 Reporting quality of articles was rated by one investigator (IF) using the STROBE checklist (STROBE Initiative. STROBE checklist for cohort studies, Version 4. University of Bern; 2007)
China
N enrolled
Author (reference), year, location
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
MEDICINE
eTABELLE 3 Subgroup analyses of included studies on education (years) and POCD (total N = 8)
Surgery type
Characteristic
Number of studies
Study, year, reference
Pooled estimates and index of heterogeneity
Moderator analysis p-value*
Cardiac surgery
5
Di Carlo 2001 (28) Hong 2008 (e37) Mathew 2007a (e6) Mathew 2007b (39) Mathew 2013 (e7)
RR: 0.90 (95% CI: [0.86; 0.95])
Versus mixed surgery type: p = 0.139
chi2(4) = 4.48; I² = 11%; p = 0.34
Versus non-cardiac surgery: p = 0.166
RR: 0.84 [0.78; 0.90]
Versus mixed surgery type: p = 0.036
Non-cardiac surgery
Sample size
chi (1) = 0.05 I² = 0%; p = 0.82
McDonagh 2010 (e1)
RR: 0.98 [0.91; 1.07]
Reference category
n ≤ 100
2
Mathew 2007a (e6) Hong 2008 (e37)
RR: 0.89 [0.80; 0.98]
p = 0.735
≤ 65 years
≤ 50% male
>50% male
Follow-up
2
1
>65 years
Sex
Monk 2008 (17) Zhu 2014 (40)
Mixed
n >100
Mean age
2
≤ 1 month
>1 month
6
4
4
3
5
2
6
Di Carlo 2001 (28) Mathew 2007b (39) Monk 2008 (17) McDonagh 2010 (e1) Mathew 2013 (e7) Zhu 2014 (40) Di Carlo 2001 (28) Mathew 2007b (39) Hong 2008 (e37) Monk 2008 (17) McDonagh 2010 (e1) Mathew 2007a (e6) Mathew 2013 (e7) Zhu 2014 (40) Hong 2008 (e37) McDonagh 2010 (e1) Monk 2008 (17) Di Carlo 2001 (28) Mathew 2007a (e6) Mathew 2007b (39) Mathew 2013 (e7) Zhu 2014 (40) Hong 2008 (e37) Zhu 2014 (40)
Di Carlo 2001 (28) Mathew 2007b (39) Mathew 2007a (e6) McDonagh 2010 (e1) Mathew 2013 (e7) Monk 2008 (17)
2
chi (1) = 0.21 I² = 0%; p = 0.65 RR: 0.91 [0.87; 0.94] chi2(5) = 12.16 I² = 59%; p = 0.03
RR: 0.87 [0.83; 0.92]
p = 0.172
2
chi (3) = 4.49; I² = 33%; p = 0.21 RR: 0.93 [0.88; 0.97] chi2(3) = 5.65 I² = 47%; p = 0.13 RR: 0.92 [0.86; 0.97])
p = 0.889
2
chi (2) = 5.99 I² = 67%; p = 0.05 RR: 0.89 [0.85; 0.94] chi2(4) = 6.11 I² = 35%; p = 0.19 RR: 0.85 [0.78; 0.92]
p = 0.159
2
chi (1)= 0.29 I² = 0%; p = 0.59 RR: 0.92 [0.88; 0.96] chi2(5) = 9.64; I² = 48%; p = 0.09
RR, relative risk per year increment in education *p-values of fixed effects moderator analyses
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
XII
MEDICINE
eFIGURE 3 SE (log[RR])
0
0.1
0.2
0.3
0.4 RR 0.5 0.5
0.7
1
1.5
2
Funnel plot for meta-analysis of years of education and POCD
eFigure 4 0
SE (log[RR])
0.5
1
1.5
RR 2 0.01
0.1
1
10
100
Funnel plot for meta-analysis of high school education versus further/higher education and POCD
XIII
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
MEDICINE
eFIGURE 5 0
SE (log[RR])
0.2
0.4
0.6
0.8 RR 1 0.02
0.1
1
10
50
Funnel plot for meta-analysis of high school education versus lower than high school education and POCD
eFigure 6 0
SE (log[RR])
0.5
1
1.5
RR 2 0.01
0.1
1
10
100
Funnel plot for meta-analysis of lower than high school education versus further/higher education and POCD
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
XIV
ORIGINAL ARTICLE
Cognitive Reserve and the Risk of Postoperative Cognitive Dysfunction A Systematic Review and Meta-Analysis Insa Feinkohl, Georg Winterer, Claudia D. Spies, Tobias Pischon
SUMMARY Background: Post-operative cognitive dysfunction (POCD) occurs in 10 to 54% of older patients during the first few weeks after surgery, but little is known about risk factors predisposing to POCD. Methods: Systematic literature review and meta-analysis of cognitive reserve indicators and POCD risk. Results: Fifteen studies on 5104 patients were included. Follow-up periods spanned 1 day to 6 months. Educational level was the most commonly assessed cognitive reserve indicator, and a longer time spent in education was associated with a reduced risk of POCD (relative risk [RR] per year increment 0.90; 95% confidence interval: [0.87; 0.94]), i.e. each year increase in education was associated with a 10% reduced risk. Similar findings were made for some analyses on education as a categorical predictor (high school versus further/higher education, RR 1.71, [1.30; 2.25]; lower than high school versus further/higher education, RR 1.69, [1.17; 2.44]) though risk was equivalent for patients with high school education and those with lower than high school education (RR 1.02; [0.78; 1.32]). Conclusion: Patients with a relatively higher level of education are at reduced risk of POCD. Risk stratification of surgical patients according to educational level may prove useful. ►Cite this as: Feinkohl I, Winterer G, Spies CD, Pischon T: Cognitive reserve and the risk of postoperative cognitive dysfunction—a systematic review and meta-analysis. Dtsch Arztebl Int 2017; 114: 110–7. DOI: 10.3238/arztebl.2017.0110
Molecular Epidemiology Research Group, Max Delbrück Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin-Buch: Insa Feinkohl, PhD, Prof. Dr. med. Pischon, MPH Charité – Universitätsmedizin Berlin: Prof. Dr. med. Winterer, Prof. Dr. med. Spies, Prof. Dr. med. Pischon, MPH MDC/BIH Biobank, Max Delbrück Center for Molecular Medicine in the Helmholtz Association (MDC), Berlin-Buch and Berlin Institute of Health (BIH), Berlin: Prof. Dr. med. Pischon, MPH
110
ost-operative cognitive dysfunction (POCD) occurs relatively frequently, in 10 to 54% of patients during the first few weeks after surgery (1). It is usually transient (2), but unlike for post-operative delirium (POD), clear diagnostic criteria are lacking for POCD (3, 4). Despite its high prevalence, POCD is underresearched and well-established risk factors for POCD are few and far between (for a review, see [2]) so that at present the cognitive risk of a surgical patient is unpredictable. Recent research has shown that diabetes (5) and pre-existing cognitive impairment (6) may predispose patients to POCD. Compared with these types of clinical risk factors, however, research into the contribution of cognitive reserve to POCD has essentially been neglected. Cognitive reserve is a theoretical construct that aims to explain links between factors such as a lower level of education, lower socioeconomic status (SES), or lower pre-morbid cognitive ability and an increased risk of cognitive impairment in older age (7–11). The account assumes that people differ in their ability to functionally ‘buffer’ neuropathological insult due to aging and disease according to their cognitive reserve capacity (12–14). Simply put, brain networks of high-reserve individuals are thought to be better able to cope with disruptions due to working more efficiently and more flexibly compared with low-reserve individuals. Neuropathological burden may further be compensated for through recruitment of novel brain networks (13). POCD is known to negatively impact on subjective cognitive function and quality of life in affected patients (15, 16). Studies suggest that it also increases the risk of dementia and mortality (17–19). POCD is thus a cause for concern from a public health perspective that exceeds problems associated with cognitive deficits alone. With a lower cognitive reserve capacity as a predictor of age-related cognitive impairment, it appears reasonable to expect an association with POCD. A lower level of education is indeed frequently discussed as a contributing factor to POCD, though empirical evidence is rarely mentioned (2, 20–23). If such evidence was to be confirmed, measures of cognitive reserve could supplement cognitive risk prediction on the basis of clinical risk factors. Because low cognitive reserve
P
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
MEDICINE
could reasonably constitute the starting point of a causal chain leading up to POCD, the identification of cognitive reserve parameters as risk factors for POCD would further add to our understanding of the processes underlying the condition. Here, we aim to integrate the current epidemiological evidence on cognitive reserve and the risk of POCD in view to providing guidance for clinical practice.
Methods Systematic search strategy An electronic search (eTable 1, eBox 1) was performed by one investigator (IF) in accordance with the MOOSE and PRISMA guidelines (24, 25). Study selection Studies were eligible for inclusion if they ● followed a prospective study design, ● included human adults undergoing surgery (age ≥18 years), ● had full texts published in English ● reported original data on associations of cognitive reserve indicators (eTable 1) with POCD in the form of odds ratios or relative risks (RR; both termed RR in the present analysis) or as descriptive data that allowed calculation of RR. Any operationalization of POCD qualified for inclusion provided it was based on performance-based neuropsychological assessment. Data extraction Fully adjusted RR statistics were extracted unless no adjustment was made. If more than one article reported on the same sample, the article with the most complete reporting was selected. Data were extracted on the longest follow-up period. For 2 studies with multivariate-adjusted data at 7 day follow-up but not at 3 months, the 7 day follow-up was selected (26, 27). For one study comparing three levels of cognitive change, ‘severe deterioration’ was used to represent POCD (28). Enquiries were made to corresponding authors for unreported information. Data synthesis and analysis Studies were analyzed separately for each cognitive reserve indicator. We used the standard I2 index to identify statistical heterogeneity (29) and inverse variance fixed-effect models to calculate summary estimates of RR (95% CI) in meta-analyses across studies. Forest plots were generated to present pooled estimates. The main meta-analyses were repeated using random-effects models (eBox 2). Potential sources of heterogeneity were explored in subgroup and metaregression analyses. Review Manager 5.3 and SAS Enterprise Guide 4.3 were used.
Results The search retrieved 109 unique articles (eFigure 1); an independent search identified a further 28 articles. Overall, 64 full text articles were assessed. Of these, Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
40 did not meet our inclusion criteria and 9 articles (30–38) were excluded due to suspected duplicate reporting (19, 26, 27, 39). In total, 15 articles were included. The included studies originated in Europe, USA, Australia, and Asia (eTable 2). Surgical procedures included cardiac surgery (n = 7) and non-cardiac surgery (n = 8) under general (n = 9) or a combination of general and regional anesthesia (n = 4) (where reported) (Table). A total of 5104 patients were analyzed. Sample characteristics varied substantially between studies. The proportion of males ranged from 26% to 79% and the mean sample age from 51 to 75 years (mean 63 ± 8 years). Patients were followed up for between 1 and 180 days after surgery (median: 25 days; interquartile range: 7 to 45 days). All studies except one (40) applied detailed batteries of neuropsychological tests, though the criteria used to define POCD were heterogeneous. POCD occurred in 8% to 67% of patients. Findings of included studies and meta-analysis a) Years of education—Eight articles reported data on years of education. The mean years of education in these studies ranged from 8 years in 2 studies from Italy and China (28, 40) to 14 years in 2 US studies (17, e1) (mean 12 ± 3 years). When effects were pooled, each year increase in education was associated with a 0.90 risk of POCD (RR 0.90 per year increment; 95% CI: [0.87; 0.94]; p16 years of education” was equated with “further/higher education.” b) High school education versus further/higher education—When effects were pooled across 4 studies (19, 27, e2, e3), high school level of education was associated with a 71% increased risk of POCD compared with a higher level of education (RR 1.71 [1.30, 2.25]; p 70
56 ± 11
6 days
1 day
7 days
3 to 30 days
3 months
RR: 2.00 [1.33; 2.98]
RR: 0.83 [0.74; 0.93]
4. RR: 2.44 [0.92; 6.50]
3. RR: 2.51 [0.90; 7.03]
1. RR: 1.47 [0.46; 4.69] 2. RR: 1.03 [0.59; 1.78]
RR: 1.01 [0.96; 1.07]
RR: 0.93 [0.85; 1.00]
Exposure association with POCD*1
1. Middle school educationversus lower than 1. RR: 1.07 [0.78; 1.32] middle school education 2. Middle school education versus further/higher 2. RR: 1.80 [0.12; 27.25] education 3. RR: 1.61 [0.11; 22.70] 3. Lower than middle school education versus further/higher education
≤16 versus >16 years of education
Years of education (mean 8 ± 3)
1. Illiterate versus literate 2. High school education versus lower than high school education 3. High school education versus further/higher education 4. Lower than high school education versus further/higher education
NART score (mean 116 ± 10)
Years of education (mean 13 ± 4)*2
69 ± 8*2 6 weeks
Exposure variable
Mean age ± standard deviation Follow-up /median (interquartile range) (years)
Data are on analysis sample unless otherwise indicated. CABG, coronary artery bypass grafting; CI, confidence interval; CR, cognitive reserve; NART, National Adult Reading Test; POCD, postoperative cognitive dysfunction; RR, relative risk; SD, standard deviation; *1definition of POCD varied between studies; see eTable 2 in Supplementary Material; *2based on total sample enrolled in the study (data on analysis sample unreported) .
316
N
Mathew et al. 2013 (e7)
Author, year (reference)
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7
MEDICINE
FIGURE a) Years of education Study (reference)
log (Relative risk)
SE
Weight
IV Fixed effects [95% CI]
Year
–0.6382 –0.1009 –0.09 –0.1381 –0.1685 –0.0161 –0.0774 –0.1863
0.2736 0.0413 0.0816 0.0674 0.0514 0.0416 0.0425 0.0586
0.5% 21.2% 5.4% 7.9% 13.7% 20.9% 20.0% 10.5%
0.53 [0.31, 0.90] 0.90 [0.83, 0.98] 0.91 [0.78, 1.07] 0.87 [0.76, 0.99] 0.84 [0.76, 0.93] 0.98 [0.91, 1.07] 0.93 [0.85, 1.01] 0.83 [0.74, 0.93]
2001 2007 2007 2008 2008 2010 2013 2014
Total [95% CI] 100.0% Heterogenity: Chi2 = 12.49. df = 7 (p = 0.09), I2 = 44% Test for overall effect: Z = 5.35 (p 16 years of education
239/440 in no-POCD group had ≤16 years of education
RR 0.83 (0.74; 0.93)
Exposure association with POCD
17/22
15/22
STROBE score*3
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
X
XI
80
Ni et al. (e3), 2015
78
N followed up
46%
% male
Intrathecal anesthesia
Total knee arthroplasty
Type of surgery and anesthesia
70 ± 4
Mean age ± SD/median (interquartile range) (years)
6 days
Follow-up
5 neuropsychological tests; MMSE
Cognitive measurement
POCD in n = 15/585 (24.8%)
Control group n = 20
POCD defined as ≥2 RCI*2 on >2 tests
Definition and incidence of POCD
1. Middle school education (n = 19/78)versus lower than middle school education (n = 56/78) 2. Middle school education versus further/ higher education (n = 3/78) 3. Lower than middle school education versus further/higher education
Exposure variable
None
Covariates
Descriptive data
Original statistical reporting
1. RR 1.07 (0.78; 1.32) 2. RR 1.80 (0.12; 27.25) 3. RR 1.61 (0.11; 22.70)
3/63 no-POCD patients with further/higher education
15/63 no-POCD patients with middle school education
45/63 no-POCD patients with lower than middle school education
0/15 POCD patients with further/higher education
4/15 POCD patients with middle school education
11/15 POCD patients with lower than middle school education
Exposure association with POCD
19/22
STROBE score*3
Data are on analysis sample unless otherwise indicated. APOEe4, apolipoprotein e4 allele; ASA class, American Association of Anesthesiology class of physical status; CABG, coronary artery bypass grafting; CR, cognitive reserve; MI, myocardial infarction; MMSE, Mini Mental Status Examination; NART, National Adult Reading Test; NYHA, New York Heart Association Functional Classification for cardiac risk; OR, odds ratio; RR, relative risk; SD, standard deviation; TMT-A, Trail-Making Test-A. *1 based on total sample enrolled in the study (data on analysis sample unreported) *2 RCI, Reliable Change Index (sometimes referred to as ‘z-scores’ in original publications). Formula to obtain RCI for each patient: (change score of patient – mean control group change score) / SD of control group change score (see [4]). *3 Reporting quality of articles was rated by one investigator (IF) using the STROBE checklist (STROBE Initiative. STROBE checklist for cohort studies, Version 4. University of Bern; 2007)
China
N enrolled
Author (reference), year, location
MEDICINE
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
MEDICINE
eTABELLE 3 Subgroup analyses of included studies on education (years) and POCD (total N = 8)
Surgery type
Characteristic
Number of studies
Study, year, reference
Pooled estimates and index of heterogeneity
Moderator analysis p-value*
Cardiac surgery
5
Di Carlo 2001 (28) Hong 2008 (e37) Mathew 2007a (e6) Mathew 2007b (39) Mathew 2013 (e7)
RR: 0.90 (95% CI: [0.86; 0.95])
Versus mixed surgery type: p = 0.139
chi2(4) = 4.48; I² = 11%; p = 0.34
Versus non-cardiac surgery: p = 0.166
RR: 0.84 [0.78; 0.90]
Versus mixed surgery type: p = 0.036
Non-cardiac surgery
Sample size
chi (1) = 0.05 I² = 0%; p = 0.82
McDonagh 2010 (e1)
RR: 0.98 [0.91; 1.07]
Reference category
n ≤ 100
2
Mathew 2007a (e6) Hong 2008 (e37)
RR: 0.89 [0.80; 0.98]
p = 0.735
≤ 65 years
≤ 50% male
>50% male
Follow-up
2
1
>65 years
Sex
Monk 2008 (17) Zhu 2014 (40)
Mixed
n >100
Mean age
2
≤ 1 month
>1 month
6
4
4
3
5
2
6
Di Carlo 2001 (28) Mathew 2007b (39) Monk 2008 (17) McDonagh 2010 (e1) Mathew 2013 (e7) Zhu 2014 (40) Di Carlo 2001 (28) Mathew 2007b (39) Hong 2008 (e37) Monk 2008 (17) McDonagh 2010 (e1) Mathew 2007a (e6) Mathew 2013 (e7) Zhu 2014 (40) Hong 2008 (e37) McDonagh 2010 (e1) Monk 2008 (17) Di Carlo 2001 (28) Mathew 2007a (e6) Mathew 2007b (39) Mathew 2013 (e7) Zhu 2014 (40) Hong 2008 (e37) Zhu 2014 (40)
Di Carlo 2001 (28) Mathew 2007b (39) Mathew 2007a (e6) McDonagh 2010 (e1) Mathew 2013 (e7) Monk 2008 (17)
2
chi (1) = 0.21 I² = 0%; p = 0.65 RR: 0.91 [0.87; 0.94] chi2(5) = 12.16 I² = 59%; p = 0.03
RR: 0.87 [0.83; 0.92]
p = 0.172
2
chi (3) = 4.49; I² = 33%; p = 0.21 RR: 0.93 [0.88; 0.97] chi2(3) = 5.65 I² = 47%; p = 0.13 RR: 0.92 [0.86; 0.97])
p = 0.889
2
chi (2) = 5.99 I² = 67%; p = 0.05 RR: 0.89 [0.85; 0.94] chi2(4) = 6.11 I² = 35%; p = 0.19 RR: 0.85 [0.78; 0.92]
p = 0.159
2
chi (1)= 0.29 I² = 0%; p = 0.59 RR: 0.92 [0.88; 0.96] chi2(5) = 9.64; I² = 48%; p = 0.09
RR, relative risk per year increment in education *p-values of fixed effects moderator analyses
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
XII
MEDICINE
eFIGURE 3 SE (log[RR])
0
0.1
0.2
0.3
0.4 RR 0.5 0.5
0.7
1
1.5
2
Funnel plot for meta-analysis of years of education and POCD
eFigure 4 0
SE (log[RR])
0.5
1
1.5
RR 2 0.01
0.1
1
10
100
Funnel plot for meta-analysis of high school education versus further/higher education and POCD
XIII
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
MEDICINE
eFIGURE 5 0
SE (log[RR])
0.2
0.4
0.6
0.8 RR 1 0.02
0.1
1
10
50
Funnel plot for meta-analysis of high school education versus lower than high school education and POCD
eFigure 6 0
SE (log[RR])
0.5
1
1.5
RR 2 0.01
0.1
1
10
100
Funnel plot for meta-analysis of lower than high school education versus further/higher education and POCD
Deutsches Ärzteblatt International | Dtsch Arztebl Int 2017; 114: 110–7 | Supplementary material
XIV
