Association Between Anxiety and Cognitive Decline Over 12 Years in a Population-Based Cohort

Abstract

Background:

Findings on the associations between anxiety and cognitive decline are mixed and often confounded.

Objective:

We studied whether anxiety symptoms were associated with the risk of cognitive decline after adequate adjustment of confounding factors.

Methods:

Our study consists of 2,551 community-dwelling older adults recruited between the ages of 60–64 years and followed up for 12 years in the PATH Through Life cohort study. Anxiety symptoms were measured using the Goldberg Anxiety Scale (GAS; range 0–9). General cognitive function, episodic memory, working memory, verbal intelligence, processing speed, and psychomotor speed were measured. Multilevel analyses were carried out to investigate the association between anxiety symptoms and cognitive decline over 12 years, taking into account confounding variables.

Results:

We did not find a significant association between baseline anxiety symptoms and cognitive decline over 12 years. Although some associations between anxiety symptoms with psychomotor speed (β= –0.04, 99% CI: –0.08, 0.00) and processing speed (β= –0.27, 99% CI: –0.48, –0.07) were found, these were attenuated after adjusting for depression. We also did not find an association between cumulative anxiety and decline in cognitive performance.

Conclusion:

In this sample of cognitively healthy men and women aged 60 years and above, anxiety symptoms were not associated with the risk of cognitive decline. Long follow-up study time, appropriate selection of confounding factors, and estimating the effect of cumulative anxiety are important to establish the association between anxiety and cognitive symptoms.

Keywords

Anxiety anxiolytics cognitive decline depression longitudinal study

INTRODUCTION

Recent failures in treating dementia have demanded a shift in the field from ‘cure’ to ‘prevention’. Delaying cognitive decline and impairment has paramount importance in limiting the number of older adults with cognitive impairment. Identification of reversible or treatable risk factors for cognitive decline is therefore of critical importance. The prevalence of anxiety in the community is high, ranging from 1.2% to 15% in cognitively intact adults aged 60 and older [1].

Anxiety may be a potentially reversible risk factor for dementia, although evidence to date has been mixed. While some cross-sectional [3] and longitudinal studies [2 , 5] show that anxiety in older adults reduces learning and memory, others [9, 10], showed no association. Correspondingly for the association between anxiety and processing speed, Gulpers et al. (2019) [5] showed that there were sex-specific differences with lower performance in women over time, while Biringer et al. (2005) showed no association [11]. These mixed results were furthermore confirmed by a systematic review which found heterogenous evidence on anxiety as a risk factor for cognitive decline, therefore limiting the scope for a meta-analysis [12].

There are several possible explanations for these contrasting findings. Firstly, diverse approaches are used to control confounding factors (e.g., depressive symptoms, comorbidities, anxiolytics, and lifestyle factors) in studying the relationship between anxiety and cognition. Increased cognitive decline in the anxious elderly could result from negative effects of the co-occurrence of depression [13, 14]. This comorbidity could be strongly influenced by the common overlap in the symptoms of depression and anxiety [15]. Long-term use of anxiolytics, especially benzodiazepines, causes cognitive impairments and should be adjusted while studying the association between anxiety and cognition [16, 17]. Along with depression and anxiety medications, anxiety is also associated with sleep disturbances, unhealthy behaviors including smoking, alcohol consumption, physical inactivity, poor diet [18, 19], and medical condition including increased risk of coronary heart disease [20]. Hence, controlling for these relevant confounders is essential to dissect the association between anxiety and cognitive decline.

Secondly, the discrepancy is related to the differences in the follow-up time of the studies. Cross-sectional studies generally support the hypothesis of a negative association between anxiety and cognitive domains in older adults [21]. However, these associations were attenuated when cognition was tested two years later [7]. Few longitudinal studies have shown that anxiety symptoms are associated with cognitive decline [5 , 23]. Studies with long follow-up time are required to better understand the relationship between anxiety and cognitive decline in older adults.

Thirdly, effects of cumulative exposure to anxiety on cognitive performance also need to be taken into consideration as it may contribute to cognitive decline. Research has shown a large number of older adults suffer from chronic anxiety [24, 25], but most epidemiological studies have focused on short-term exposure and geared mainly to baseline measurements of anxiety. This is in line with depression, where persistent episodes of depression are associated with dementia and cognitive decline, but single isolated episodes of depression do not have a strong effect [26]. To understand the association between anxiety and cognition it is also important for studies to distinguish anxiety symptoms from those meeting a diagnostic threshold (clinically significant anxiety) [27]. This is particularly important as this may be another reason for inconsistent findings found in the literature. We, therefore, identify a need for more prospective studies to help clarify the association between clinical anxiety and cognitive decline/dementia.

We hypothesize that the lack of conformity of the association between anxiety and cognitive measures are mainly due to inadequate adjustments of the confounding factors, including depression, use of anxiolytics, lifestyle factors, and medical factors. Moreover, lack of long follow-up time and inadequate consideration of chronicity of anxiety are other possible parameters that need to be tested for association. To address these issues in the current study, we assessed the association between anxiety symptoms and different cognitive functions in a well-characterized large cohort of older adults dwelling in Australia, with follow-ups over 12 years. We adjusted for a number of diverse and well-known confounders in a step-by-step process to analyze the association between anxiety and each cognitive domain. We also assessed whether the cumulative exposure to anxiety significantly reduced the cognitive function over time.

METHODS

Participants

Data were obtained from the 60 s cohort of the PATH Through Life study, an ongoing longitudinal cohort study of health and aging originally situated in Canberra/Queanbeyan, Australia [28]. At baseline, participants aged between 60 to 64 years on 1 January 2001 were randomly recruited from the electoral roll of the Australian Capital Territory and Queanbeyan in Australia. Voting is compulsory in Australia. Comparison with Australian census data show the final cohort is representative of the general population in terms of marital and employment status but have higher education status (for further demographic details, see Anstey et al. [28]).This study included data from the first four waves of the study collected at 4-year intervals over 12 years. 2,551 participants were recruited at baseline. The following exclusions were made based on baseline data: 134 participants scoring less than 27 on the Mini-Mental State Examination (MMSE), 26 participants diagnosed with MCI or dementia, and 8 participants with missing baseline anxiety measurements. Of the 2,383 participants measured at baseline, 2,113 (88.7%) and 1,888 (79.2%) participants were available in wave 2 and wave 3 respectively. Finally, 1,578 (66.2% of the initial sample) were assessed in wave 4. Details on the cohort profile for wave 4 along with attrition and missing data in PATH cohort are reported elsewhere [29, 30]. The study was approved by the Australian National University Human Research Ethics Committee, and all participants gave written informed consent.

Cognitive measures

A battery of cognitive measures was used. Immediate and delayed recall which measures episodic memory was assessed using the first trial of the California Verbal Learning Test (CVLT) [31]. To minimize practice effects, an alternate word-list was used for immediate recall for the 60 s cohort at wave 4 [32], but was otherwise identical in administration to prior waves. At wave 4, the full version of the CVLT was administered. This was not comparable to the earlier waves and hence it was not included. Spot the word Test Version A was carried out in the 60 s cohort to assess premorbid, verbal intelligence [33]. Wechsler Memory Scale-Digit Span Backwards was used to assess working memory [34]. Symbol Digit Modalities Test (SDMT) assessed processing speed [35]. The Purdue Pegboard test is a cognitive-motor task used to assess psychomotor speed [36]. The MMSE, a dementia screening instrument was also used as a measure of global cognitive impairment [37].

Anxiety assessment

Anxiety symptoms were measured using the 9-item anxiety scale (GAS) of the Goldberg Anxiety and Depression scale (GADS) [38]. This tool measures anxiety symptoms over the past four weeks through binary (yes/no) responses. The GADS score was calculated using the positive responses to these items. Along with the continuous GAS score, we also categorized participants based on the severity of anxiety into the following three groups: no anxiety (0), mild anxiety (1–4 GAS score), and moderate/severe anxiety (≥5 GAS score). As Goldberg and colleagues, recommend a cut-off score of 5 suggests that a participant has a 50% chance of “clinically important disturbance” of anxiety, while higher scores substantially increase the probability of a significant anxiety disorder [38]. Finally, we estimated the level of chronicity of anxiety (cumulative anxiety) by counting the number of waves participants had moderate/severe anxiety.

Confounders

Possible confounders that may be associated with cognition measurements and anxiety include socio-demographics, lifestyle, and medical conditions. The socio-demographic variables age, sex, years of education, and non-English speaking background (NESB) and financial hardship were included as possible confounders. Financial hardship was a good measure of socioeconomic status obtained at each wave as a self-reported question on whether the person had to go without basic needs being met in the past year due to financial hardship (financial hardship: yes/no) [29, 39]. Lifestyle factors which include smoking status (current, former, never), alcohol consumptions, body mass index (BMI: normal, overweight, and obese), physical activity, and social support were also adjusted for as confounders. Alcohol consumption was assessed using the Alcohol Use Disorders Identification Test (AUDIT) [40]. Alcohol consumption (calculated according to National Health and Medical Research Council 2001 guidelines [41]) using number of drinks per week with <0.25/week as being abstained and light to moderate intake in males being 0.25–20.5 drinks per week and in females being 0.25–13.5 drinks per week. Physical activity was calculated as a combination of self-reported number of hours/week performing mild, moderate, and vigorous activities, weighted by 1, 2, and 3, respectively and subsequently categorized as low, moderate and high intensities [42]. Social support was a composite measure of positive and negative social exchange with family and friends [43].

Medical variables including current depression (measured by Goldberg Depression Scale), insomnia, Framingham cardiovascular risk scores, physical health (measured by SF-12 survey), sleep medication (yes/no response to whether medications for sleep were taken in the last month), anxiolytics (yes/no response to taking any medications for anxiety), and use of benzodiazepines (taking benzodiazepines for any reason which included Valium, Xanax, Mogadon and whether taking for anxiety, depression, or insomnia purpose) were also included as confounders. All the analyses were also adjusted for APOE genotypes.

Statistical analyses

The basic demographic and key study characteristics were presented using mean, standard deviation, count, and proportion where appropriate. The association between the baseline anxiety score and cognitive measurements across all waves were examined using linear mixed effect models with a subject specific random intercept. All covariates were included as a fixed term in the model. To examine the role of depression, anxiety medication along with other confounders, we assessed all the confounders using four different hierarchical models in the following manner. The first Model (Model 1) assessed the association between cognitive measures and baseline anxiety symptoms by controlling for the effect of baseline confounders such as age, sex, education, NESB, the wave of data collection, and APOE genotypes. Depression symptoms, along with all the covariates that were adjusted in Model 1, were included in Model 2. Model 3 adjusted for depression and anxiolytics along with all the baseline covariates from Model 1. Finally, Model 4 adjusted for all the covariates in Model 3 plus a selection of significant covariates for each of the cognitive measures as the outcome. Additional baseline covariates were selected for inclusion in the multivariable models by backward elimination of the least significant variable until all variables were less than p < 0.05 using Likelihood ratio test.

To understand the effect of cumulative anxiety on the change in cognitive functions, we used multivariate linear regression models adjusting for baseline cognitive measurements and other baseline confounders in the similar hierarchical fashion described above. In all the above models, a 100(1–0.05/7) ≈ 99% CI was calculated to allow for the multiple testing correction using the Bonferroni method. For the assessment of the impact of missing data in covariates and outcome, we revised all the analysis using multiple imputed dataset with fully conditional specifications [44]. The regression coefficients were averaged over 30 imputed datasets.

All analyses were conducted using Stata version 16.0 (StataCorp, College Station, TX).

RESULTS

Baseline characteristics

The baseline characteristics of the study sample are presented according to the severity of anxiety in Table 1. At baseline, we found more than 50% of the study sample had mild anxiety symptoms while 18% were categorized as having moderate to severe symptoms. A high number of participants with moderate to severe anxiety symptoms smoked, practiced unhealthy lifestyle habits, were more depressed, and had other medical conditions as compared to those with mild or no anxiety symptoms. Participants with moderate to severe anxiety symptoms performed slightly lower on the tests for global cognition, processing speed, and working memory. At baseline, a very small proportion of data were missing for alcohol consumption, smoking, sleep and anxiety medication, physical health, and social support score (see Supplementary Table 1). In terms of longitudinal missing data, there was no difference in age, sex, and baseline anxiety levels on comparing lost to follow-up participants with those who were interviewed at wave 4 (see Supplementary Table 2). The drop out group had slightly lower education, a higher percentage of non-English speaking background participants and generally scored lower on cognitive measures at baseline.

Table 1

Characteristics of study sample according to the anxiety level at baseline

Characteristics	No anxiety symptoms n = 725	Mild (1–4 anxiety symptoms) n = 1,229	Moderate to severe (> 5 anxiety symptoms) n = 429
Age, y, mean±SD	62±1.5	63±1.5	62±1.5
Female n (%)	305 (42.1)	616 (50.1)	251 (58.5)
Race n (%)
Caucasian	691 (95.3)	1189 (96.8)	415 (96.7)
Asian	24 (3.3)	18 (1.5)	8 (1.9)
Other	10 (1.4)	22 (1.8)	5 (1.2)
No of years of education M (p25-p75)	14 (13–16)	14 (12–16)	13 (11–16)
Non-English speaking background (NESB) n (%)	42 (5.8)	83 (6.8)	30 (7.0)
APOE ɛ4 code
ɛ4-/ɛ4-	507 (74.0)	841 (72.5)	791 (72.4)
ɛ4-/ɛ4+	168 (24.5)	289 (24.9)	104 (25.9)
ɛ4 + /ɛ4+	10 (1.5)	30 (2.6)	7 (1.7)
Missing	40 (5.5)	69 (5.6)	27 (6.3)
Goldberg depression score
M (p25-p75)	0 (0-1)	1 (0–2)	4 (2–5)
Alcohol group, n (%)
Light/moderate	602 (83.3)	995 (81.1)	318 (74.3)
Abstain	78 (10.8)	163 (13.3)	80 (18.7)
Heavy drinkers	43 (6.0)	69 (5.6)	30 (7.0)
Missing	2 (0.3)	2 (0.2)	1 (0.2)
Smoking status n (%)
Never smoker	401 (55.3)	634 (51.6)	204 (47.7)
Former smoker	257 (35.5)	465 (37.8)	172 (40.2)
Current smoker	67 (9.2)	130 (10.6)	52 (12.2)
Exercise n (%)
Low	271 (43.4)	606 (54.7)	244 (64.5)
Moderate	253 (40.5)	352 (31.8)	105 (27.8)
High	101 (16.2)	149 (13.5)	29 (7.7)
Missing	100 (13.8)	122 (9.9)	51 (5.9)
BMI n (%)
Normal	279 (41.5)	425 (38.4)	159 (41.7)
Overweight	277 (41.2)	477 (43.1)	133 (34.9)
Obese	117 (17.4)	204 (18.4)	89 (23.4)
Missing	52 (7.2)	123 (10.0)	48 (11.2)
Trouble sleeping n (%)
Not at all	642 (88.6)	720 (58.6)	87 (20.8)
Several days	80 (11.0)	380 (30.9)	193 (45.0)
More than half the days	1 (0.1)	66 (5.4)	81 (18.9)
Nearly everyday	2 (0.3)	63 (5.1)	68 (15.8)
Anxiety medication use n (%)	9 (1.2)	42 (3.4)	73 (17.0)
Sleep medication n (%)	30 (4.1)	157 (12.8)	131 (30.6)
Framingham risk score mean±SD	0.67±0.24	0.68±0.25	0.70±0.27
Physical health score M (p25-p75)	56 (52–56)	52 (44–56)	44 (32–51)
Schuster-social support score median (p25-p75)	15 (8–20)	12 (7–17)	7 (2–13)
Cognitive measures M (p25-p75)
Immediate recall	7 (6–9)	7 (6–9)	7 (6–9)
Delayed recall	6 (5–8)	6 (5–8)	6 (5–8)
Spot-the-word	53 (49–56)	53 (49–56)	53 (49–56)
Symbol digit modality test	51 (47–57)	51 (45–56)	50 (43–53)
MMSE	30 (29–30)	30 (29–30)	29 (29–30)
Perdue pegboard	11 (9–12)	11 (9–12)	10 (9–12)
Digit backward	5 (3–7)	5 (3–7)	4 (3–6)

Note: M indicates median, p25 and p75 are the 25th and 75th percentiles.

Association between anxiety symptoms and cognitive functions over 12 years

Table 2 shows the associations between anxiety symptoms and cognitive function over 12 years. In the basic model (Model 1) when adjusted for age, sex, race, APOE genotypes, education, and NESB, anxiety symptoms were significantly associated with only Perdue Pegboard score (β= –0.04, 99% CI: –0.08, 0.00) and SDMT score (β= –0.27 99% CI: –0.48, –0.07). However, after adjusting for depressive symptoms (Model 2) and other covariates (Model 3 and 4), these associations were no longer significant. Details on the models of various cognitive outcome measures are presented in Supplementary Tables 4 to 10. We have also analyzed the interaction effect of anxiety by time in linear mixed model with random intercept and slopes over different time period (refer Supplementary Table 14). The results are similar to those with random intercept model. Moreover, a likelihood-ratio test between random intercept and slope model and a random intercept only model suggested that there is no significant association between anxiety and cognitive function over time (results not shown).

Table 2

Multilevel model estimates (99% CI) for the association between anxiety and cognitive measures

	Immediate recall	Delayed recall	MMSE	Spot the word	Perdue pegboard	Digit backward	SDMT
Model 1	–0.01	–0.02	–0.01	0.05	–0.04	–0.01	–0.27
	(–0.05, 0.03)	(–0.07, 0.03)	(–0.03, 0.01)	(–0.06, 0.16)	(–0.08, –0.00)	(–0.06, 0.04)	(–0.48, –0.07)
Model 2	0.02	0.00	0.00	0.02	–0.01	0.01	–0.04
	(–0.03, 0.08)	(–0.06, 0.07)	(–0.02, 0.03)	(–0.12, 0.16)	(–0.06, 0.04)	(–0.05, 0.07)	(–0.31, 0.22)
Model 3	0.03	0.01	0.01	0.01	0.00	0.01	–0.02
	(–0.02, 0.08)	(–0.05, 0.07)	(–0.02, 0.03)	(–0.13, 0.16)	(–0.05, 0.05)	(–0.13, 0.16)	(–0.29, 0.25)
Model 4	0.03	0.01	0.01	0.03	0.03	0.11
	(–0.02, 0.09)	(–0.04, 0.06)	(–0.01, 0.04)	(–0.11, 0.17)	(–0.02, 0.08)	(–0.16, 0.38)

Model 1: Adjusted for age, sex, education, NESB, APOE ɛ4, financial hardship, and wave of data collection. Model 2: Adjusted for depression along with all the covariates in model 1. Model 3: Adjusted for depression and anxiolytics along with all the covariates in model 1. Model 4: Adjusted for all the covariates in model 3 plus a selection of significant covariates for each outcome selected using backward elimination (p < 0.05) for each of the outcome. For a) Immediate recall and delayed recall: alcohol, sleeping problem and physical heath; b) MMSE and SDMT: alcohol and physical heath; c) Spot the word: alcohol, smoking status and physical heath, d) Perdue pegboard: BMI, Physical health and Framingham risk score, e) Digit backward: no additional covariates are significant.

The association between anxiety symptoms and cognitive functions over 12 years using an imputed dataset are presented in Supplementary Table 11. Overall, these results are very similar with complete case analysis with none of the cognitive measures except the SDMT score (β= –0.26 99% CI: –0.46, –0.05) showing association with baseline anxiety measures.

Cumulative anxiety and its association with cognitive change over 12 years

In addition to the association between anxiety symptoms and longitudinal measures of cognitive function over time, we also examined the effect of cumulative anxiety on the cognitive change between wave 1 and wave 4. Overall, 79% (1,887/2,383) of participants did not have any anxiety at any of the four waves, whereas 13% (316) participants had moderate/severe anxiety in only one wave, 4% (102) in two waves, 2% (56) in three waves and 1% (22) in all four waves (Supplementary Table 3). Table 3 shows the effect of cumulative anxiety on the cognitive change between wave 1 and wave 4. We did not find an association between cumulative anxiety and decline in cognitive performance. However, participants with short-term moderate/severe anxiety symptoms (GAS≥5 in one wave of the four waves) was associated with change in working memory. This association was significant in the basic model (Model 1: β= –0.40 99% CI: –0.79, –0.01) and remained significant after adjusting for depression (Model 2: β= –0.41 99% CI: –0.81, –0.01), depression and use of anxiolytics (Model 3: β= –0.41 99% CI: –0.81, –0.01), and depression, use of anxiolytics, and social support (Model 4: β= –0.44 99% CI: –0.84, –0.03). Similar results were also obtained when regrouping chronic anxiety as 0 waves with anxiety, 1 wave with anxiety, 2+ waves with anxiety (Supplementary Table 12). However, all of these associations were no longer significant in the multiple imputation datasets (see Supplementary Table 13).

Table 3

Multivariate linear regression estimates (99% CI) for the association between numbers of waves with anxiety and the change in cognition measurements over the 12-year period

	No of waves with anxiety	Immediate recall	Delayed recall	MMSE	Spot the word	Perdue pegboard	Digit backward	SDMT
Model 1	0	Ref	Ref	Ref	Ref	Ref	Ref	Ref
	1	–020 (–0.55, 0.15)	–0.16 (–0.54, 0.23)	–0.15 (–0.46, 0.16)	0.11 (–0.44, 0.67)	–031 (–0.72, 0.10)	–0.40 (–0.79, –0.01)	–1.30 (–2.73, 0.13)
	2	–0.31 (–0.88, 0.25)	–0.52 (–1.14, 0.10)	–0.26 (–0.77, 0.25)	0.50 (–0.40, 1.41)	–0.23 (–0.90, 0.44)	–0.12 (–0.74, 0.51)	–0.30 (–2.61, 2.01)
	3	–0.28 (–1.00, 0.45)	0.34 (–0.43, 1.12)	0.07 (–0.59, 0.73)	0.31 (–0.86, 1.48)	–0.18 (–1.05, 0.69)	–0.40 (–1.24, 0.44)	–0.78 (–3.75, 2.20)
	4	–0.24 (–1.28, 0.80)	–0.09 (–1.23, 1.05)	0.03 (–0.97, 1.03)	–0.52 (–2.36, 1.32)	–0.03 (–1.41, 1.36)	–0.40 (–1.61, 0.81)	–1.74 (–6.23, 2.75)
Model 2	0	Ref	Ref	Ref	Ref	Ref	Ref	Ref
	1	–0.16 (–0.53, 0.20)	–0.12 (–0.53, 0.28)	–0.11 (–0.44, 0.21)	0.20 (–0.38, 0.78)	–0.24 (–0.67, 0.19)	–0.41 (–0.81, –0.01)	–0.93 (–2.42, 0.55)
	2	–0.23 (–0.85, 0.34)	–0.45 (–1.11, 0.20)	–0.18 (–0.72, 0.36)	0.66 (–0.29, 1.61)	–0.10 (–0.81, 0.61)	–0.13 (–0.79, 0.53)	0.40 (–2.04, 2.84)
	3	–0.19 (–0.97, 0.53)	0.41 (–0.40, 1.22)	0.14 (–0.54, 0.82)	0.45 (–0.75, 1.66)	–0.07 (–0.96, 0.82)	–0.42 (–1.28, 0.45)	–0.15 (–3.20, 2.91)
	4	–0.12 (–1.23, 0.94)	0.02 (–1.18, 1.21)	0.16 (–0.88, 1.20)	–0.23 (–2.15, 1.69)	0.18 (–1.25, 1.62)	–0.42 (–1.68, 0.84)	–0.52 (–5.21, 4.17)
Model 3	0	Ref	Ref	Ref	Ref	Ref	Ref	Ref
	1	–0.16 (–0.53, 0.21)	–0.12 (–0.53, 0.28)	–0.11 (–0.44, 0.22)	0.21 (–0.37, 0.79)	–0.24 (–0.67, 0.19)	–0.41 (–0.81, –0.01)	–0.91 (–2.39, 0.57)
	2	–0.25 (–0.83, 0.37)	–0.45 (–1.10, 0.21)	–0.14 (–0.69, 0.40)	0.70 (–0.26, 1.65)	–0.08 (–0.79, 0.63)	–0.14 (–0.80, 0.52)	0.55 (–1.89, 2.99)
	3	–0.22 (–0.94, 0.56)	0.42 (–0.40, 1.23)	0.22 (–0.46, 0.90)	0.52 (–0.69, 1.74)	–0.04 (–0.93, 0.85)	–0.42 (–1.29, 0.44)	0.16 (–2.91, 3.23)
	4	–0.15 (–1.21, 0.97)	0.02 (–1.18, 1.22)	0.22 (–0.82, 1.26)	–0.17 (–2.09, 1.75)	0.22 (–1.22, 1.66)	–0.43 (–1.70, 0.83)	–0.29 (–4.98, 4.40)
Model 4	0		Ref		Ref	Ref	Ref	Ref
	1		–0.09 (–0.49, 0.32)		0.21 (–0.40, 0.81)	–0.12 (–0.58, 0.34)	–0.44 (–0.84, –0.03)	–1.00 (–2.48, 0.49)
	2		–0.41 (–1.07, 0.25)		0.69 (–0.30, 1.69)	0.11 (–0.63, 0.89)	–0.13 (–0.79, 0.53)	0.40 (–2.04, 2.85)
	3		0.44 (–0.37, 1.26)		0.59 (–0.67, 1.84)	0.06 (–0.89, 1.01)	–0.53 (–1.42, 0.37)	0.04 (–3.03, 3.11)
	4		0.07 (–1.13, 1.26)		–0.05 (–1.98, 1.89)	0.15 (–1.31, 1.61)	–0.49 (–1.76, 0.77)	–0.87 (–5.56, 3.82)

Model 1: Adjusted for age, sex, education, NESB, and APOE ɛ4 genotype at baseline. Model 2: Adjusted for depression at baseline along with all the covariates from model 1. Model 3: Adjusted for depression and anxiolytics used at baselines along with all the covariates in model 1. Model 4: Adjusted for all the covariates in model 3 plus a selection of significant covariates for each outcome selected using backward elimination (p < 0.05) for each of the outcome. For a) Immediate recall and MMSE: no additional confounders were selected; b) delayed recall: physical heath; c) Spot the word: BMI at baseline, d) Perdue pegboard: BMI, Physical health and Framingham risk score, e) Digit backward: social support score, and f) SDMT: sleep problem and Framingham risk score.

DISCUSSION

This study examined the association between baseline anxiety symptoms and cognitive functions over 12 years in a large sample of community-living elderly persons. We found no association between anxiety symptoms and decline in memory, verbal intelligence, processing speed, psychomotor speed, and global cognition after adjusting for depression. Moreover, we did not find an association between cumulative anxiety and a decline in cognitive performance. Anxiety is believed to be a probable modifiable risk factor for cognitive decline, although evidence to date has been mixed. Results from our study do not support this notion as we did not find an association between anxiety and cognitive decline in older adults.

We found the relationship between anxiety symptoms and cognitive performance was strongly influenced by depressive symptoms. Individuals with anxiety symptoms performed poorly in Perdue Pegboard and SDMT, indicating reduced psychomotor and processing speed, but these associations were altered after adjusting for depression as well as in multiple imputed datasets. Our findings are similar to those of de Bruijn et al. [10] and Biringer et al. [11], who showed that anxiety does not contribute to cognitive decline as the negative effects of anxiety on processing speed were mostly explained by co-morbid depression. A recent study [45] has also shown that comorbid anxiety symptoms have no additive effect on cognitive dysfunction in late-life depression. These results highlight the importance of distinguishing the unique contribution of anxiety symptoms from depressive symptoms on cognition.

In our study, we did not find an association between cumulative anxiety and decline in cognitive performance. These analyses are important as they help segregate the difference in effect of anxiety symptoms from clinically significant anxiety (individuals who meet a diagnostic threshold). A systematic review by Gimson et al. (2018) [27] showed that clinically significant anxiety at midlife was positively associated with a high risk of dementia, leading them to conclude that the association of midlife clinically significant anxiety and late-life dementia is as strong as that between late-life anxiety symptoms and dementia. To understand the association between anxiety and cognition, future studies need to clearly distinguish anxiety symptoms versus clinical anxiety in relation to cognitive decline and dementia risk.

Although in our study we do not find any associations between anxiety and cognitive decline, several plausible mechanisms underlying the relationship between anxiety and cognitive processes are identified in the literature. 1) Anxiety may act as a modifiable risk factor for faster cognitive decline operating through various biomedical or lifestyle factors. Cortisol levels are known to increase with stress and anxiety. Excessive cortisol is secreted through the activation of the hypothalamus-pituitary-adrenal axis, which in turn activates glucocorticoid receptors that are present in memory-forming structures, hippocampus, and prefrontal cortex. Both these structures are involved in memory and higher-order executive functioning and are sensitive to changes in levels of cortisol induced by stress [47]. While we did not use cortisol measure in this analysis, it would be interesting to check the levels of cortisol in the anxious participants over time and to determine its possible mediating or moderating role in the context of cognition functions. 2) Anxiety could be a prodromal syndrome of oncoming cognitive impairment; while our study was a community-dwelling population, several other studies included participants with mild cognitive impairment. For example, Palmer et al. [48] have shown that among participants with mild cognitive impairment, anxiety was a predictor for conversion to dementia whereas it was not for cognitively healthy subjects. 3) Anxiety and cognitive decline may appear to be linked due to some common cause (e.g., education, genetics). Brain derived neurotrophic factor and Apolipoprotein E ɛ4 are further examples of genetic factors that are common to both anxiety and cognitive decline [49, 50].

Our findings should be placed in the context of the strengths and limitations of the study. A particular strength of the current study is the study design. Our study is a community-based study with a large number of older adults, with the availability of a comprehensive set of potentially confounding covariates and measures of different cognitive domains. In addition, our study had a long follow-up time of 12 years, which is required to clarify the association between anxiety and cognitive decline. Furthermore, and most importantly, we thoroughly examined the association between anxiety symptoms and cognition, taking into consideration a large number of relevant confounders including depression, the use of anxiolytics, lifestyle habits including smoking, alcohol consumption, obesity, and various medical factors. This is not always the case in published studies. This study also has limitations that include self-report of anxiety symptoms without clinical assessment and the 4-year gap between the self-reports. In our analysis, cumulative anxiety was measured by the number of waves (not necessarily adjacent waves) participants had exposures of moderate/severe anxiety over a period of time. However, chronic anxiety would specifically measure consecutive captures of moderate/severe anxiety over the waves. Further analyses need to be carried out to investigate the association between chronic anxiety and cognitive function. Lastly, we cannot generalize our results due to the relatively well educated Australian urban population which was largely Caucasian.

CONCLUSION

Our study on cognitively healthy older men and women shows that anxiety symptoms are not associated with a decline in cognitive functions. These findings suggest that to establish the relationship between anxiety and cognition, long follow-up study time, along with adequate adjustment for confounders including depression, the use of anxiolytics and other related confounders are crucial.

Footnotes

ACKNOWLEDGMENTS

The authors thank the PATH study participants and acknowledge the PATH interviewers and study team for their contributions.

We thank the study participants, PATH interviewers, project team, and contributions of Tony Jorm, Helen Christensen, Bryan Rodgers, Keith Dear, Simon Easteal, Peter Butterworth, and Nicolas Cherbuin. KJA is supported by National Health and Medical Research Council (NHMRC) Fellowship (No. 1002560) and acknowledges support from the Australian Research Council Centre of Research Excellence in Population Ageing Research (CE110001029). MEM is supported by the NHMRC and Australian Research Council (ARC) Dementia Research Development Fellowship (No. 1102028). The PATH Through Life Study was funded by NHMRC Grants (No. 973302, 179839, 418039, 1002160).

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Bryant

, Jackson

, Ames

(2008) The prevalence of anxiety in older adults: Methodological issues and a review of the literature. J Affect Disord 109, 233–250.

Bierman

EJM

, Comijs

, Jonker

, Beekman

ATF

(2005) Effects of anxiety versus depression on cognition in later life. Am J Geriatr Psychiatry 13, 686–693.

Yochim

, Mueller

, Segal

(2013) Late life anxiety is associated with decreased memory and executive functioning in community dwelling older adults. J Anxiety Disord 27, 567–575.

Okereke

, Grodstein

(2013) Phobic anxiety and cognitive performance over 4 years among community-dwelling older women in the Nurses’ Health Study. Am J Geriatr Psychiatry 21, 1125–1134.

Gulpers

BJA

, Oude Voshaar

, van Boxtel

MPJ

, Verhey

FRJ

, Kohler

(2019) Anxiety as a risk factor for cognitive decline: A 12-year follow-up cohort study. Am J Geriatr Psychiatry 27, 42–52.

Moran

(2016) Anxiety and working memory capacity: A meta-analysis and narrative review. Psychol Bull 142, 831–864.

Brodaty

, Heffernan

, Draper

, Reppermund

, Kochan

, Slavin

, Trollor

, Sachdev

(2012) Neuropsychiatric symptoms in older people with and without cognitive impairment. J Alzheimers Dis 31, 411–420.

Kassem

, Ganguli

, Yaffe

, Hanlon

, Lopez

, Wilson

, Cauley

(2017) Anxiety symptoms and risk of cognitive decline in older community-dwelling men. Int Psychogeriatr 29, 1137–1145.

Bierman

, Comijs

, Rijmen

, Jonker

, Beekman

(2008) Anxiety symptoms and cognitive performance in later life: Results from the longitudinal aging study Amsterdam. Aging Ment Health 12, 517–523.

10.

de Bruijn

, Direk

, Mirza

, Hofman

, Koudstaal

, Tiemeier

, Ikram

(2014) Anxiety is not associated with the risk of dementia or cognitive decline: The Rotterdam Study. Am J Geriatr Psychiatry 22, 1382–1390.

11.

Biringer

, Mykletun

, Dahl

, Smith

, Engedal

, Nygaard

, Lund

(2005) The association between depression, anxiety, and cognitive function in the elderly general population–the Hordaland Health Study. Int J Geriatr Psychiatry 20, 989–997.

12.

Gulpers

, Ramakers

, Hamel

, Kohler

, Oude Voshaar

, Verhey

(2016) Anxiety as a predictor for cognitive decline and dementia: A systematic review and meta-analysis. Am J Geriatr Psychiatry 24, 823–842.

13.

Pietrzak

, Kinley

, Afifi

, Enns

, Fawcett

, Sareen

(2013) Subsyndromal depression in the United States: Prevalence, course, and risk for incident psychiatric outcomes. Psychol Med 43, 1401–1414.

14.

Breitve

, Hynninen

, Bronnick

, Chwiszczuk

, Auestad

, Aarsland

, Rongve

(2016) A longitudinal study of anxiety and cognitive decline in dementia with Lewy bodies and Alzheimer’s disease. Alzheimers Res Ther 8, 3.

15.

Mineka

, Watson

, Clark

(1998) Comorbidity of anxiety and unipolar mood disorders. J Ann Rev Psychol 49, 377–412.

16.

Picton

, Marino

, Nealy

(2018) Benzodiazepine use and cognitive decline in the elderly. Am J Health Syst Pharm 75, e6–e12.

17.

Paterniti

, Dufouil

, Alpérovitch

(2002) Long-term benzodiazepine use and cognitive decline in the elderly: The Epidemiology of Vascular Aging Study. J Clin Psychopharmacol 22, 285–293.

18.

Mehta

, Simonsick

, Penninx

, Schulz

, Rubin

, Satterfield

, Yaffe

(2003) Prevalence and correlates of anxiety symptoms in well-functioning older adults: Findings from the health aging and body composition study. J Am Geriatr Soc 51, 499–504.

19.

Strine

, Chapman

, Kobau

, Balluz

(2005) Associations of self-reported anxiety symptoms with health-related quality of life and health behaviors. Soc Psychiatry Psychiatr Epidemiol 40, 432–438.

20.

Nabi

, Hall

, Koskenvuo

, Singh-Manoux

, Oksanen

, Suominen

, Kivimaki

, Vahtera

(2010) Psychological and somatic symptoms of anxiety and risk of coronary heart disease: The health and social support prospective cohort study. Biol Psychiatry 67, 378–385.

21.

Beaudreau

, O’Hara

(2008) Late-life anxiety and cognitive impairment: A review. Am J Geriatr Psychiatry 16, 790–803.

22.

Bunce

, Batterham

, Mackinnon

, Christensen

(2012) Depression, anxiety and cognition in community-dwelling adults aged 70 years and over. J Psychiatr Res 46, 1662–1666.

23.

Gallacher

, Bayer

, Fish

, Pickering

, Pedro

, Dunstan

, Ebrahim

, Ben-Shlomo

(2009) Does anxiety affect risk of dementia? Findings from the Caerphilly Prospective Study. Psychosom Med 71, 659–666.

24.

Schuurmans

, Comijs

, Beekman

, de Beurs

, Deeg

, Emmelkamp

, van

Dyck R

(2005) The outcome of anxiety disorders in older people at 6-year follow-up: Results from the Longitudinal Aging Study Amsterdam. Acta Psychiatr Scand 111, 420–428.

25.

de Beurs

, Beekman

, Deeg

, Van Dyck

, van Tilburg

(2000) Predictors of change in anxiety symptoms of older persons: Results from the Longitudinal Aging Study Amsterdam. Psychol Med 30, 515–527.

26.

Zheng

, Zhong

, Song

, Xie

(2018) Persistent depressive symptoms and cognitive decline in older adults. Br J Psychiatry 213, 638–644.

27.

Gimson

, Schlosser

, Huntley

, Marchant

(2018) Support for midlife anxiety diagnosis as an independent risk factor for dementia: A systematic review. BMJ Open 8, e019399.

28.

Anstey

, Christensen

, Butterworth

, Easteal

, Mackinnon

, Jacomb

, Maxwell

, Rodgers

, Windsor

, Cherbuin

, Jorm

(2012) Cohort Profile: The PATH through life project. Int J Epidemiol 41, 951–960.

29.

Anstey

, Peters

, Mortby

, Kiely

, Eramudugolla

, Cherbuin

, Huque

, Dixon

(2021) Association of sex differences in dementia risk factors with sex differences in memory decline in a population-based cohort spanning 20-76 years. Sci Rep 11, 7710.

30.

Anstey

, Butterworth

, Christensen

, Easteal

, Cherbuin

, Leach

, Burns

, Kiely

, Mortby

, Eramudugolla

, Gad

(2021) Cohort Profile Update: The PATH Through Life Project. Int J Epidemiol 50, 35–36.

31.

Delis

, Kramer

, Kaplan

, Ober

(1987) California Verbal Learning Test, Psychological Corporation, Harcourt Brace Jovanovich, San Antonio.

32.

Delis

, Massman

, Kaplan

, Mckee

, Kramer

, Gettman

(1991) Alternate form of the California Verbal Learning Test: Development and reliability. Clin Neuropsychol 5, 154–162.

33.

Baddeley

, Emslie

, Nimmo-Smith

(1993) The Spot-the-Word test: A robust estimate of verbal intelligence based on lexical decision. Br J Clin Psychol 32, 55–65.

34.

Wechsler

(1945) A standardized memory scale for clinical use. J Psychol 19, 87–95.

35.

Smith

(1982) Symbol digit modalities test, Western Psychological Services, Los Angeles.

36.

Tiffin

(1968) Purdue Pegboard examiner manual. Science Research Associates, Chicago, IL.

37.

Molloy

, Alemayehu

, Roberts

(1991) Reliability of a standardized mini-mental state examination compared with the traditional mini-mental state examination. Am J Psychiatry 148, 102–105.

38.

Goldberg

, Bridges

, Duncan-Jones

, Grayson

(1988) Detecting anxiety and depression in general medical settings. BMJ 297, 897–899.

39.

Butterworth

, Rodgers

, Windsor

(2009) Financial hardship, socio-economic position and depression: Results from the PATH Through Life Survey. Soc Sci Med 69, 229–237.

40.

Saunders

, Aasland

, Babor

, de la Fuente

, Grant

(1993) Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with Harmful Alcohol Consumption–II. Addiction 88, 791–804.

41.

NHMRC (2001) Australian alcohol guidelines: Health risks and benefits. National Health and Medical Research Council.

42.

Singh-Manoux

, Hillsdon

, Brunner

, Marmot

(2005) Effects of physical activity on cognitive functioning in middle age: Evidence from the Whitehall II prospective cohort study. Am J Public Health 95, 2252–2258.

43.

Schuster

, Kessler

, Aseltine

Jr (1990) Supportive interactions, negative interactions, and depressed mood. Am J Community Psychol 18, 423–438.

44.

Huque

, Carlin

, Simpson

, Lee

(2018) A comparison of multiple imputation methods for missing data in longitudinal studies. BMC Med Res Methodol 18, 1–16.

45.

Martinussen

, Saltyte Benth

, Almdahl

, Borza

, Selbaek

, McPherson

, Korsnes

(2019) The effect of anxiety on cognition in older adult inpatients with depression: Results from a multicenter observational study. Heliyon 5, e02235.

46.

Mantella

, Butters

, Dew

, Mulsant

, Begley

, Tracey

, Shear

, Reynolds

, 3rd Lenze

(2007) Cognitive impairment in late-life generalized anxiety disorder. Am J Geriatr Psychiatry 15, 673–679.

47.

Arnsten

(2009) Stress signalling pathways that impair prefrontal cortex structure and function. Nat Rev Neurosci 10, 410–422.

48.

Palmer

, Berger

, Monastero

, Winblad

, Bäckman

, Fratiglioni

(2007) Predictors of progression from mild cognitive impairment to Alzheimer disease. Neurology 68, 1596–1602.

49.

Michels

, Multhammer

, Zintl

, Mendoza

, Klünemann

(2012) Association of apolipoprotein E ɛ4 (ApoE ɛ4) homozygosity with psychiatric behavioral symptoms. J Alzheimers Dis 28, 25–32.

50.

Suliman

, Hemmings

, Seedat

(2013) Brain-derived neurotrophic factor (BDNF) protein levels in anxiety disorders: Systematic review and meta-regression analysis. Front Integr Neurosci 7, 55.