Decision-Making Profiles and Their Associations with Cognitive Performance in Mild Cognitive Impairment

Abstract

Background:

It is crucial for older adults, especially those with mild cognitive impairment (MCI), to make profitable decisions routinely. However, the results regarding decision-making (DM) remain inconsistent.

Objective:

The present study assessed DM profiles under uncertainty conditions in individuals with MCI and their associations with multi-domain cognitive performance.

Method:

Fifty-three patients with MCI and forty-two age-, gender-, and education level-matched healthy controls (HCs) were administered a comprehensive neuropsychological battery test. The Iowa Gambling Task (IGT) and Game of Dice Task (GDT) were used to assess DM competence in conditions involving ambiguity and risk, respectively. In addition, Spearman’s correlations were used to examine relationships between GDT and multi-domain cognitive performance.

Result:

The final capital (FC) and frequency of utilization of negative feedback (FUNF) and positive feedback (FUPF) in the GDT were lower in MCI patients than in HCs. In addition, the number of shifts between safe and risky alternatives was significantly different across groups. However, IGT performance was comparable across groups. In the MCI patients, risky DM performance was associated with language, whereas in HCs was correlated with memory and executive functions. Besides, in MCI, performance on IGT was significantly correlated with social cognition.

Conclusion:

Individuals with mild cognitive impairment have difficulty utilizing feedback to make optimal decisions under risky situations. The association between decision-making performance and cognitive function is divergent regarding situational uncertainty and individuals’ cognitive status. In mild cognitive impairment and normal aging, decision-making under ambiguity needs further investigation.

Keywords

Ambiguous situation cognitive function decision-making mild cognitive impairment risky situation situational uncertainty

INTRODUCTION

Mild cognitive impairment (MCI) is a clinical stage on the continuum of cognitive decline, which involves no significant decline in the functional activities of daily living. Its presence is associated with a higher risk of dementia [1]. Therefore, early detection of MCI is desperately needed. Although MCI shows intact daily functioning, deficits may arise in more complex situations, such as decision-making (DM) related to finance, medical issues, driving, or consent to treatment [2 –4].

DM is a complex process in which individuals select between multiple options associated with uncertain consequences [5]. There are two types of DM primarily based on the type and amount of available information. Probabilities of different outcomes and possible rewards and punishments are given or calculated in risky conditions, whereas information about outcome probabilities is missing or conflicting in ambiguous conditions [6, 7]. The Iowa Gambling Task (IGT), a clinically sensitive tool that measures DM in ambiguous situations, and the Game of Dice Task (GDT), assessing a patient’s capacity to make decisions under risk, are well established. These two instruments have been widely applied in the context of neurological and psychiatric disorders [7, 8].

DM is highly relevant to everyday functioning and autonomy. Furthermore, compelling evidence has suggested that DM deficits may be a harbinger of conversion to dementia [9]. Indeed, DM seems highly sensitive to cognitive changes [10, 11]. Therefore, the assessment of DM in individuals with MCI may be helpful in the early detection of individuals with a higher risk of developing dementia. However, there is little known about the characteristics of DM in MCI. Although a few studies have explored whether individuals with MCI show difficulties in making favorable decisions in situations involving ambiguity and risk, the nature and extent of DM dysfunction associated with this illness remain conflicting [12 –17]. Some studies have described the worse performance of patients with MCI relative to healthy controls (HCs) under both ambiguous and risky conditions, which is similar to findings with individuals with mild Alzheimer’s disease [12 , 18]. However, other studies suggested that the frequency of feedback utilization in GDT and the performance in IGT were comparable among MCI and HCs [15].

These contradictory results may have resulted from the fact that the measurement outcomes were too simple to recognize the pattern of DM [15]. Previous studies have shown that the IGT’s two most different card decks were more sensitive than the original net score [19]. In addition, due to the limited sample size, the results of previous studies may not have efficiently demonstrated DM performance in individuals with MCI [12 , 16]. Consequently, the present study aimed to examine the characteristics of DM impairments in conditions with ambiguity and risk.

In recent years, the relationship between DM and cognitive function has been reported, such as attention, episodic memory, visuospatial ability, and aspects of executive functioning [12 , 20–25]. Therefore, it is plausible that cognitive impairment may be associated with poorer DM capacity. However, the existing literature has generated inconsistent findings. Several studies found that DM performance under both risky and ambiguous conditions significantly correlated with executive functions, attention, and memory [21, 23]. Others suggested a significant correlation between either ambiguous DM or risky DM but not both and social cognitive function, attention, executive function, and memory [12 , 26]. Therefore, diverse relationships have been shown to exist regarding DM under risk or ambiguity and cognitive functions in individuals with MCI. As most previous studies combined data from patient and control groups, it remained unknown whether these associations were disease specific [12 , 16]. Also, when examining cognitive performance, most existing literature treated the measures of neuropsychological tests separately. The collective effect of the tests tapping into specific domains, which might be measured with the composite score [27], was not explored extensively. Therefore, to have better insight into the potential cognitive correlates of DM, it is necessary to thoroughly examine the performance in major cognitive domains.

In the present study, our principal aim was to determine the characteristics of DM abilities in conditions of uncertainty, including risky and ambiguous situations. Another goal of this study was to establish whether these competencies were significantly correlated with deficits in specific cognitive domains. We speculated that MCI patients would have difficulties making good choices and recognizing feedback in uncertain conditions, especially in the ambiguous situation. The correlations between DM deficits and cognitive domains performance would differ between MCI patients and HCs.

METHODS

Participants

Ninety-five participants, including 53 individuals with MCI and 42 HCs, were recruited at the Dementia Care and Research Center, Peking University Institute of Mental Health, between August 2019 and June 2021. All participants completed a standardized neuropsychological assessment, clinical interview, and magnetic resonance imaging (MRI) examination and received a clinical diagnosis by senior memory specialists. Furthermore, 52 participants with MCI and 42 HCs completed the GDT, 52 with MCI and 39 HCs completed the IGT.

Eligible patients were equal to or older than 60 years and had a diagnosis of MCI according to Petersen’s criteria [28], which included the following: memory complaint confirmed by an informant; normal general cognitive function, as defined by global Clinical Dementia Rating (CDR) score of 0.5 [29] and a Mini-Mental State Examination (MMSE) score between 24 and 30 [30]; preserved activities of daily living, as defined by Functional Activities Questionnaire≤6 [31]; and a lack of dementia, according to the International Classification of Diseases, 10th Revision (ICD-10) [32] and the criteria for probable AD of the National Institute of Neurological and Communicative Disorders and the Stroke/Alzheimer Disease and Related Disorders Association (NINCDS-ADRDA) [33]. In addition, all subjects had primary school education (≥5 years); scored four or below on the apathy subdomain score of the Neuropsychiatric Inventory[34]; had a Hachinski Ischemic Score < 4, and brain MRI showing no other diseases capable of producing cognitiveimpairment.

The inclusion criteria for HCs were as follows: 1) were aged at least 60 years, 2) had no memory complaints, 3) had more than five years of education, 4) did not have detected cognitive impairment, indicated by an MMSE score≥24, Montreal Cognitive Assessment (MoCA) score≥26 [35], and a CDR score = 0.

Participants were excluded if they had a history of other psychiatric disorders (e.g., schizophrenia, bipolar disorder, major depressive disorder, alcohol or substance abuse/dependence), stroke, other focal brain injury or other neurodegenerative condition, any hearing or visual impairment that could prevent an efficient evaluation, been taking any drugs able to affect cognition or a Geriatric Depression Scale score of 11 or higher [36].

The Ethics Board of Peking University Institute of Mental Health approved the present study. Written informed consent was obtained from all participants.

Neuropsychological tests

All patients performed the Chinese Neuropsy-chological-Normative Consensus Battery (CNCB; CNNORM Consensus Battery) [37], which is acomprehensive neuropsychological battery that includes attention (digit span-forward, trail making test-A), memory (Hopkins verbal learning test, brief visual memory test), executive function (trail making test-B, digit span-backward, digit symbol substitution, Stroop color-word test), language (verbal fluency (animal naming), Boston naming test), visuospatial (clock drawing test, visual object, and space perception-silhouettes), and social cognition (reading the mind in the Eyes).

Except for the social cognition domain, raw scores on the 12 tests mentioned above were converted to z-scores using means and standard deviations. Five corresponding composite scores for five cognitive domains (attention, memory, executive function, language, and visuospatial ability) were created by averaging the z-scores of the measures within a domain, as previously reported [38, 39].

Decision-making under risk

The GDT is a computerized task that measures DM under risk. As previously described by Sun et al. [15], participants were instructed to choose dice indicating the gain or loss of the momentary capital in a fictitious gambling game. Before the test, the rules for gains/losses associated with each alternative were explicitly described to all participants. The game consisted of 18 trials, starting from the capital of ¥ 1000. The participants could choose the dice to bet on any single number or combinations of two, three, or four numbers in each trial. Each alternative was associated with a specific amount of fictive gain or loss based on a specific winning probability. The amounts of gains or losses, the residual capital, and the number of remaining trials were visible on the screen throughout the test. From the beginning of the task, the participants could calculate the winning probability associated with each alternative and apply feedback from previous trials to maximizegains.

Choices and strategy changes in the GDT were used as indicators of decision-making under risk in our study. Choices in the GDT were determined as 1) final capital (FC); 2) the number of selections from each of the four possible alternatives; 3) the frequency of choosing risky (selection of single-number and two-number alternatives) and safe (selection of three-number and four-number alternatives) options; and 4) net scores, which were the result of subtracting the frequency of choosing risky options from the frequency of choosing safe options. Strategy changes in the GDT were defined as 1) the net scores in trials 1 to 6, 7 to 12, and 13 to 18 to examine selection patterns across groups; 2) the frequency of utilizing negative feedback (FUNF; the frequency of trials in which a participant shifted to select safe options after negative feedback) and positive feedback (FUPF; the frequency of trials in which a participant continued to select safe options after positive feedback); and 3) the number of shifts between safe and risky alternatives (NSBSR) to assess feedbackeffects.

Decision-making under ambiguity

The IGT is a computer-based tool widely used to examine ambiguous decision-making capacity [40, 41]. In this test, each card choice was linked with different probabilities and amounts of gain and loss, while the participants were not instructed with the probabilities and amounts explicitly. Briefly, The participants were asked to choose a card from one of the four different decks (A, B, C, D) to maximize the fictional capital starting from ¥ 2000 loan of play money in 100 trials: Decks A and B (defined as disadvantageous decks) were high immediate gain (¥ 100/choice), but more significant cumulative penalties that would lead to a negative balance in the long term; Decks C and D (defined as advantageous decks) were associated with low immediate gain (¥ 50/choice) but more minor cumulative penalties that participants would accumulate incremental profit in the long term. After each choice, the patient could receive feedback on the amount of money won or lost in the single trial and the cumulativecapital.

In our study, decision-making capacity in ambiguous conditions was measured with the choices made during the 100 trials and strategy changes in IGT. The choices measures included the FC, the number of card selections from each of the four decks, the total number of card selections from the disadvantageous decks (A and B), and the total number of selections from the advantageous decks (C and D), and the difference in selecting the cards of low immediate gain, less probable but more significant penalty (deck D) and those of high immediate gain, highly probable but smaller penalty (deck A), i.e., the net score of (D-A) [19]. In addition, strategy changes in IGT were defined by the net scores for every 20 selected cards calculated in the same manner to examine changes in performance throughout the test, the utilization of negative (FUNF) and positive (FUPF) feedbacks, and the number of shifts between safe and risky alternatives(NSBSR).

Sample size estimation

In reference to a previous study [15], the power of 0.80 and the significant level of 0.05 were set. The sample size was estimated to be N = 37 for each group.

Statistical analysis

The Chi-square test was conducted to assess group differences in the gender distribution. In addition, age, education level, and cognitive characteristics, normally distributed, were analyzed with independent t-tests to assess group differences. Finally, other variables were analyzed by nonparametric Mann–Whitney U tests. The effect size of group comparisons was also calculated.

We used the generalized estimating equation (GEE) to examine the variables of choices in the GDT and IGT. The grouping variable was included as the between-subject factor. In addition, the order of trials (trials 1–6, trials 7–12, trials 13–18) and blocks (block 1–5) was included as the within-subject factor for the analysis of the GDT and IGT net scores, respectively. Besides, the frequency of penalty (high or low) was also included as the within-subject factor in the GEE model when we compared the number of card selections from each frequency penalties cards (B + D or A + C) in the IGT between MCI and HC groups.

The Spearman rank-order correlation analysis was used to explore the relationship between decision-making measures and the scores of each cognitive domain. The composite z score was used for attention, memory, executive function, language, and visuospatial ability domains. Moreover, the raw score of reading the mind in eyes was used for social cognition.

All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) software version 23.0. All 2-sided p < 0.05 was considered to be statistically significant. Measures of GDT and IGT were divided into two subsets: making choice and change of strategy. When group differences were significant in the subsets of data in GDT and IGT, Bonferroni correction was applied. Similarly, when the correlation between cognitive performance and performance of decision-making was significant, Bonferroni correction was also applied for multiple correlation analyses.

RESULTS

Demographic and clinical characteristics of the sample

The characteristics of the participants are provided in Table 1. Gender (χ² = 0.023; p = 0.879), years of education (t = 1.314, p = 0.192), and age (t = –0.190; p = 0.061) were not significantly different between groups. The scores of global cognitive functions were significantly greater in the HCs group than the MCI patients (MMSE: Z = –6.47, p < 0.001; MoCA: Z = –8.37, p < 0.001). Except for the attention z-score (attention: Z = –0.959; p = 0.337), the other five specific z-transformed cognitive outcomes (memory: Z = –7.569; p < 0.001; executive functions: Z = –4.541, p < 0.001; language: Z = –5.654; p < 0.001; visuospatial ability: Z = –4.991, p < 0.001) and the raw score of social cognition (Z = –3.783; p < 0.001) were significantly lower in the MCI patients than in the HCs. The raw scores of each cognitive domain were described in Supplementary Table 1.

Table 1

Demographic and clinical characteristics of the participants

	MCI (n = 53)	HCs (n = 42)	χ²/t/Z	Effect size (Cohen’ d / r)^¶	p
Female^*	32 (60.4%)	26 (61.9%)	0.02	0.0	0.879
Education (y)^**	14.66±1.84	15.17±1.90	1.31	–0.3	0.192
Age (y)^**	74.91±6.31	72.33±6.85	–0.19	0.4	0.061
MMSE^***	27 (26, 29)	30 (29, 30)	–6.47	0.7	< 0.001
MoCA^***	22 (20, 23)	28 (27, 29)	–8.37	0.9	< 0.001
Attention z-score^***	–0.12 (–0.70, 0.32)	0.07 (–0.51, 0.34)	–0.96	0.1	0.337
Memory z-score^***	–2.21 (–2.81, –1.66)	0.33 (–0.82, 0.72)	–7.57	0.8	< 0.001
Executive Functions z-score^***	–0.59 (–0.76, –0.28)	–0.01 (–0.38, 0.22)	–4.54	0.5	< 0.001
Language z-score^***	–1.26 (–1.88, –0.43)	0.05 (–0.38, 0.40)	–5.65	0.6	< 0.001
Visuospatial ability z-score^***	–1.34 (–2.16, 0.03)	0.03 (–0.25, 0.37)	–4.99	0.5	< 0.001
Social Cognition^*** (Reading the Mind in the Eyes (emotion))	23 (19, 25)	26 (23, 28)	–3.78	0.4	< 0.001

^*n (%); ^**mean (standard deviation); ^***median (interquartile range); ^¶r was used to measure effect size for nonparametric tests. MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; MCI, mild cognitive impairment; HCs, healthy controls.

Decision-making under risk

In the GDT task, the individuals with MCI gained significantly less than the HCs measured by the FC (–900 versus 700; Z = –2.252, p = 0.024). The subjects with MCI utilized negative feedback (Z = –2.359, p = 0.018) and positive feedback (Z = –2.171, p = 0.03) less often than the HCs. In addition, those with MCI shifted more frequently between risky and safe alternatives (Z = –2.836, corrected p = 0.005, Table 2). Other GDT measures were not significantly different between the MCI and HC groups (p > 0.05).

Table 2

Comparison of performance on the Game of Dice Task (GDT) in the MCI and HCs groups

	MCI (n = 52)	HCs (n = 42)	Z	Effect size (r)	p
Single-number alternative	3.00 (1.00, 5.75)	0.00 (0.00, 8.00)	–1.76	0.2	0.078
Double-number alternative	3.00 (1.00, 5.75)	2.00 (1.00, 4.00)	–1.36	0.2	0.174
Triple-number alternative	4.00 (2.25, 6.00)	3.00 (1.00, 5.25)	–1.78	0.2	0.075
Quadruple-number alternative	5.50 (1.25, 8.75)	5.00 (1.00, 13.25)	–1.09	0.1	0.276
Safe alternative	10.50 (7.00,15.00)	13.50 (4.00, 17.00)	–0.97	0.1	0.331
Net score_GDT	3.00 (–4.00, 2.00)	9.00 (–10.00, 16.00)	–0.97	0.1	0.331
FC	–900 (–4725, 775)	700 (–2825, 1400)	–2.25	0.2	0.024
FUNF	1.00 (1.00, 2.00)	2.00 (1.00, 2.25)	–2.36	0.3	0.018
FUPF	4.00 (2.00, 7.50)	8.00 (3.00, 9.00)	–2.17	0.2	0.030
NSBSR	4.00 (2.00, 7.00)	2.00 (1.00, 4.00)	–2.84	0.3	0.005

The continuous variables in the table are the number of each alternative selected in the GDT. Values are shown as medians (interquartile ranges). GDT, Game of Dice Task; MCI, mild cognitive impairment; HCs, healthy controls; FC, final capital; FUNF, frequency of utilization of negative feedback; FUPF, frequency of utilization of positive feedback; NSBSR, number of shifts between safe and risky alternatives.

In the GEE model of the net scores of the GDT, the main effect of trials was significant (Wald χ² = 21.97; p < 0.001). Nevertheless, the main effects of group (Wald χ² = 0.414, p = 0.520) and trial^*group were not significantly different (Wald χ² = 2.281; p = 0.320). As shown in Fig. 1a, post hoc analysis indicated that the net scores in trials 7 to 12 (Wald χ² = 22.68, p < 0.001) and in trials 13 to 18 (Wald χ² = 11.179, p = 0.001) were significantly higher than the net scores in trials 1 to 6. However, there was no difference between trials 7 to 12 and 13 to 18 (Wald χ² = 0.897, p = 0.343).

Fig. 1

Violin plots depict the distribution of the net scores of the first, second, and last third trials of GDT (trials 1 to 6, 7 to 12, and 13 to 18, respectively) in MCI and HC groups. The violin plot indicated that the distribution of net scores in the first was significantly less than in the second and last third trials. Moreover, there was no statistical difference between the distribution of net scores in the second and last third trials. The solid lines in the violin plots represented the median. Furthermore, dotted lines in the violin plots represented the first quartile and third quartile. The width of the violin plots describes how frequently that value occurs in the data set. The broader regions of the plots indicated values that occur more frequently. Narrower ones indicated that the value occurs less frequently. MCI, mild cognitive impairment; HCs, healthy controls; GDT, Game of Dice Task.

Regarding the preference for each alternative in GDT, only the main effect of alternatives was significant (Wald χ² = 15.087; p = 0.002). The main effects of group (Wald χ² = 2.168, p = 0.520) and alternative^*group were not significantly (Wald χ² = 3.038; p = 0.386). A post hoc analysis indicated that MCI and HC groups both preferred the quadruple-number alternative relative to the single-, double-, and triple-number alternative (Wald χ² = 5.881, p = 0.015, Wald χ² = 13.974, p < 0.001, Wald χ² = 6.706, p = 0.01, respectively).

Decision-making under ambiguity

All IGT measures did not significantly differ between the MCI and HC groups (see details in Table 3). The GEE model revealed a significant learning effect from the first to the fifth block of the IGT (Wald χ² = 18.21; p < 0.001). However, the learning effect was similar in the two groups (Wald χ² = 0.69, p = 0.406), and there was no significant group-by-block interaction (Wald χ² = 2.67, p = 0.445) (Fig. 2). Besides, both groups preferred to choose decks associated with low-frequency penalties (B + D) rather than those with high-frequency penalties (A + C) (Wald χ² = 119.132, p < 0.001, Fig. 3).

Table 3

Comparison of the performance on the Iowa gambling task (IGT) in the MCI and HCs groups

	MCI (n = 52)	HCs (n = 39)	Z	Effect size (r)	p
Deck A	20 (15, 24)	19 (15, 24)	–0.37	0.0	0.715
Deck B	31 (25, 36)	30 (24, 40)	–0.71	0.1	0.478
Deck C	21 (18, 26)	21 (17, 25)	–0.73	0.1	0.467
Deck D	29 (23, 35)	29 (23, 34)	–0.11	0.0	0.910
FC	1588 (1163, 1969)	1500 (1300, 1750)	–0.47	0.1	0.642
Original net score (C + D)-(A + B)	2 (18)	2(24)	–0.59	0.1	0.555
Modified net score (D-A)	7.5 (–1, 19)	9 (3, 14)	–0.22	0.0	0.828
Advantageous choices	51 (46, 55)	51 (43, 55)	–0.59	0.1	0.555
FUNF	6 (4, 8)	6 (3, 8)	–0.30	0.1	0.765
FUPF	20 (15, 30)	20 (14, 27)	–0.911	0.1	0.362
NSBSR	44 (30, 55)	46 (31, 56)	–0.57	0.1	0.569

The continuous variables in the table are the number of cards selected in decks. Values are shown as medians (interquartile ranges). IGT, Iowa gambling task; MCI, mild cognitive impairment; HCs, healthy controls; FC, final capital; FUNF, frequency of utilization of negative feedback; FUPF, frequency of utilizing positive feedback; NSBSR, number of shifts between safe and risky alternatives.

Fig. 2

IGT Performance in MCI and HC groups. Each block (1–5) represents 20 sequential card choices. The net score was calculated by subtracting the number of deck D selections from deck A selections. Triangles and squares represent the median for each subject group in each block. MCI, mild cognitive impairment; HCs, healthy controls; IGT, Iowa gambling task.

Fig. 3

The violin plots depict the distribution of low-frequency penalties (B + D) cards and high-frequency penalties (A + C) over 100 trials in the IGT. The violin plots revealed the distribution of choices in the IGT, which indicated that decks with low-frequency penalties (B + D) were significantly more than the high-frequency penalties (A + C) ones in MCI and HCs. The broader regions of the violin plots indicated that the deck was selected more frequently. Narrower ones indicated that the deck was selected less frequently. MCI, mild cognitive impairment; HCs, healthy controls; IGT, Iowa gambling task.

Association between decision making performance and cognitive abilities

As shown in Table 4, the relationship between decision-making performance and cognitive abilities was divergent in the MCI and HC groups. In the MCI group, there was a modest correlation between the GDT measures, including the net score, choosing safe alternatives, and utilizing positive feedback (FUPF), and the composite score of language (r = 0.375∼0.395, corrected p < 0.01). Better language performance was associated with more winning strategies in the GDT testing. In the IGT test, the use of negative feedback (FUNF) correlated significantly with the performance of the reading mind in the eyes (r = 0.386, corrected p < 0.01), indicating that better social cognition may be associated with more use of winning strategies in the IGT tasks (Table 5). In addition, the frequency of advantageous choices (r = 0.304, uncorrected p = 0.028) and modified net score (r = 0.329, uncorrected p = 0.017) were marginally correlated with social cognition (Table 5).

Table 4

Correlations between GDT performance and specific cognitive abilities in the MCI and HCs groups

	MCI						HCs
	Attention	Memory	Executive Functions	Language	Visuospatial	Social Cognition	Attention	Memory	Executive Functions	Language	Visuospatial	Social Cognition
Net score_GDT	0.072	0.202	0.112	0.375 ^*	0.160	0.102	0.202	0.492 ^*	0.471 ^*	0.081	0.342	0.124
Single-number alternative	–0.099	–0.149	–0.13	–0.279	–0.251	–0.083	–0.176	–0.397 ^*	–0.482 ^*	–0.105	–0.332	–0.085
Double-number alternative	–0.244	–0.034	0.022	–0.237	0.101	0.001	–0.219	–0.267	–0.112	0.173	<0.001	–0.097
Triple-number alternative	–0.027	0.264	0.055	0.148	0.267	0.137	0.037	0.139	0.148	0.091	0.311	–0.026
Quadruple-number alternative	0.002	0.022	0.100	0.220	0.069	0.096	0.172	0.427 ^*	0.529 ^*	0.067	0.169	0.100
Safe alternative	0.072	0.202	0.112	0.375 ^*	0.160	0.102	0.202	0.492 ^*	0.471 ^*	0.081	0.342	0.124
FC	–0.039	0.249	0.210	0.347	0.291	0.154	0.089	0.378	0.461 ^*	0.079	0.232	0.073
FUNF	–0.013	–0.155	0.139	–0.253	0.081	0.043	0.038	0.140	0.29	0.047	0.223	0.061
FUPF	0.116	0.162	–0.032	0.395 ^*	0.061	0.031	0.172	0.386	0.439 ^*	–0.073	0.012	–0.017
NSBSR	–0.203	–0.040	0.08	–0.246	0.196	0.057	–0.174	–0.206	0.000	0.010	0.023	–0.040

Among the MCI patients MCI, 52 participants completed the GDT, whereas 50 and 48 patients used negative and positive feedback, respectively. Among the HCs, 42 participants completed the GDT, whereas 33 and 37 used negative and positive feedback. R values are presented. GDT, Game of Dice Task; MCI, mild cognitive impairment; HCs, healthy controls; FC, final capital; FUNF, frequency of utilization of negative feedback; FUPF, frequency of utilization of positive feedback; NSBSR, number of shifts between safe and risky alternatives. ^* Corrected p < 0.01.

Table 5

Correlations between IGT performance and specific cognitive abilities in the MCI and HCs groups

	MCI						NC
	Attention	Memory	Executive Functions	Language	Visuospatial	Social Cognition	Attention	Memory	Executive Functions	Language	Visuospatial	Social Cognition
Deck A	0.158	–0.171	–0.040	–0.178	–0.077	–0.190	0.093	0.194	0.275	0.317	0.247	–0.094
Deck B	0.16	0.022	–0.147	–0.157	–0.118	–0.252	–0.025	–0.089	–0.021	–0.201	0.049	–0.115
Deck C	–0.071	0.091	–0.007	0.141	–0.123	0.082	–0.110	–0.157	–0.271	–0.138	–0.344	0.092
Deck D	–0.252	0.078	0.165	0.115	0.224	0.332	–0.021	0.083	–0.014	0.049	0.024	0.164
Advantageous choices	–0.171	–0.004	0.149	0.227	0.154	0.304	–0.055	–0.056	–0.228	–0.100	–0.216	0.198
Modified net score (D-A)	–0.238	0.135	0.136	0.192	0.181	0.329	–0.040	–0.082	–0.189	–0.112	–0.080	0.162
FC	–0.257	0.206	0.084	–0.032	0.088	0.196	–0.225	–0.049	–0.130	–0.055	–0.158	0.026
FUNF	0.152	–0.149	0.01	0.076	0.225	0.386 ^*	0.265	–0.068	0.013	–0.102	0.073	–0.037
FUPF	–0.082	–0.024	0.168	0.079	–0.019	0.213	–0.111	–0.133	–0.197	–0.229	–0.408	–0.052
NSBSR	–0.128	0.177	–0.094	0.168	0.291	0.189	0.128	0.103	0.045	0.121	0.288	0.193

Among the MCI patients MCI, 52 participants completed the IGT. Among the HCs, 42 participants completed the IGT. R values are presented. IGT, Iowa gambling task; MCI: mild cognitive impairment; HCs, healthy controls; FC, final capital; FUNF, frequency of utilization of negative feedback; FUPF, frequency of utilizing positive feedback; NSBSR, number of shifts between safe and risky alternatives. ^*Corrected p < 0.01.

In contrast, in the HC group, there was a significant moderate correlation between the dice selection pattern in the GDT, measured with the net score, single-number alternative, quadruple-number alternative and safe alternative, and the composite scores of the memory performance and executive function (value of r ranged from 0.397 to 0.529, all corrected p < 0.01, Table 4). Better memory and executive function were associated with more gaining choices. Also, the FC and FUPF were significantly correlated with executive function (r = 0.439∼0.471, corrected p < 0.01, Table 4). Better executive function was associated with more gains and winning strategies in the decision under risky conditions. However, in the IGT task, the frequency of selecting deck C (r = –0.344, uncorrected p = 0.032) and FUPF (r = –0.408, uncorrected p = 0.01) were marginally correlated with the visuospatial ability (Table 5). There was no significant relationship between IGT measures and other cognitive measures.

DISCUSSION

Our study was one of the first studies that examined the performance of decision-making under risky and ambiguous situations among individuals with mild cognitive impairment and its relationship with major domains of cognitive function. Briefly, the significant findings included: 1) Individuals with MCI performed poorer than normal controls in decision-making under risk: gaining less measured with final capital; less often using feedbacks to make advantage choices, and shifting more often when selecting risky or safe alternatives. 2) The decision-making performance under ambiguity was similar in MCI and normal controls. 3) GDT performance correlated with language ability in MCI and with memory and executive function in healthy controls. In contrast, the IGT performance correlated with social cognition in MCI and was marginally related to visuospatial function in healthy controls.

In the GDT, the patients with MCI had more difficulties making decisions than the HCs. The finding was consistent with previous studies in patients with MCI [12 , 15]. In the present study, the net score and the frequency of choosing the four possible alternatives, risky options and safe options in risk, were similar in MCI patients and HCs. However, this study showed that feedback utilization and the number of shifts between safe and risky alternatives were significantly different between MCI patients and HCs under risky conditions. The ability to process feedback might be deficient in patients with MCI under risky situations. Thus, being unable to recognize positive or negative feedback led to MCI patients having problems selecting an advantageous option or continuing to choose a disadvantageous option after receiving feedback under precise conditions. In other words, the patients with MCI preferred to make risky decisions without considering the consequences of their behaviors in GDT. In a pilot study, Sun et al. did not find significant differences in negative feedback utilization and shifting between advantageous and disadvantageous selections in the GDT task between MCI and controls[15]. Compared with the previous report (effect size = 0.2 for the PUNF in GDT), our study (effect size = 0.3 for the PUNF in GDT) may take advantage of a larger sample size allowing sufficient power to detect the group difference.

Furthermore, in the present study, the lower levels of feedback utilization and shifting more frequently between risky and safe choices in the patients with MCI may have impaired risky DM function. A review of studies revealed that successful DM under risk depended on an intact and complex neural network, such as the dorsolateral prefrontal cortices (DLPFC), orbitofrontal cortex, inferior frontal gyrus, caudate, and anterior cingulate cortex (ACC) [7]. Dysfunction of any of these brain regions may contribute to changes in DM. Previous neuroimaging studies showed that MCI individuals exhibited slight but significant disconnection among the DLPFC, orbitofrontal cortex, inferior frontal gyrus, caudate, and ACC [42 –44]. Therefore, we could speculate that the brain structural and functional alterations in MCI may contribute to the vulnerability to decision-making failures.

Surprisingly, the present study did not observe a significant difference in the IGT measures between MCI and controls. The finding was consistent with a previous study [15]. Previous studies on DM in MCI under ambiguous conditions have yielded inconsistent results [12 , 17]. Zamarian et al. suggested that MCI patients performed more poorly than HCs in block 5 of the IGT [12]. Bayard et al. found that the net score in blocks 3–5 of the IGT in amnestic MCI patients was significantly lower than in HCs [16]. Although the IGT is a tool widely used to examine ambiguous decision competence, whether later stages of the IGT measure ambiguous DM remains a controversial topic. Also, the engagement in social and leisure activities, and neuropsychiatric symptoms, such as apathy, might have influenced DM in amnestic MCI patients [7 , 45]. Besides, in our study, there was a high frequency of shifts when selecting advantageous or disadvantageous decks and a low frequency of consistently favorable options in both groups. Therefore, the performance might result in an insignificant difference between the two groups. A previous study indicated that IGT performance depended on conscious knowledge about the task. It also claimed that the IGT might be too complex for older adults to be conscious of the task rules [46]. Furthermore, only 37.5% –52.5% of healthy adults indicated a clear preference for advantageous over disadvantageous decks in the IGT [47, 48]. It might partly explain the findings of decision-making under ambiguity in our study.

Our results went a step further to indicate that both groups preferred the two decks that yielded infrequent penalties (B and D) over those that yielded more frequent penalties (A and C). There were no significant differences between the MCI patients and HCs. There is increasing evidence to suggest that subjects’ performances are not affected by punishment frequency but by the magnitude of losses and gains [49]. Previous studies investigated sensitivity to punishment frequency in AD patients. Compared with HCs, AD patients selected deck B or deck D [15, 19]. To our knowledge, this is the first study to explore whether the punishment frequency might influence choice behavior in patients with MCI under ambiguity. The present study results indicated that both groups appeared to be sensitive to the probability rather than magnitude of losses and gains, which resulted in similar choice strategies in both groups. The capacity to integrate simultaneous information about outcome frequency and magnitude was poor in the elderly population, including those with MCI or HCs, under ambiguous conditions, which could explain why this study failed to find a general difference in the frequency of selecting deck B and deck D between patients with MCI and HCs.

In addition, we found that successful DM under risky conditions in MCI might depend on good language ability. At the same time, it relied on intact memory and executive functions in the healthy control. Successful performance in GDT might be linked with executive functions when participants are required to categorize safe and risky alternatives, develop and apply strategies, and integrate feedback [13, 50]. Furthermore, memory might be necessary for forming and updating the alternatives associated with rewards and punishments in good decision-making under risk. Our finding that language correlated with DM performance in MCI under risky conditions was interesting. A recent study reported that MCI individuals presented mild deficits in language processing [51]. In our study, however, the participants were clearly instructed and confirmed to have understood the instructions by trained investigators. Thus, MCI individuals in the present study might not discover the hidden messages of the feedback. Thus, it may partly account for the poor performance in making good selections and utilizing positive or negative feedback.

Moreover, the present study revealed that the better social cognition, the better decision-making in MCI patients in IGT. Previous studies suggested that the divergent association of brain regions linked with ambiguity and risky decision-making might reflect possible differential functionality [52]. Successful performance on the IGT was strongly associated with activity in the ventromedial and orbitofrontal prefrontal cortices and the amygdala [7]. Furthermore, the ventromedial prefrontal cortex mediated social cognition measured by reading the mind in the eyes task [53]. The pathological lesions might result in the relationships between social cognition and IGT performance in MCI.

In contrast, the IGT performance of healthy controls correlated with visuospatial abilities. The mechanism of visuospatial abilities and decision-making performance was unclear. As discussed earlier, the IGT task may be too complex to process for older adults. Under the ambiguous situation, individuals may bring as many resources as possible to sort out the probability of gains or penalties of the card choices. As there was no explicit message about each choice, the individual may prefer to count on integrating visual information rather than memory and executive function. However, the neural mechanism of deciding on ambiguous situations warrants further investigation.

Limitations

The findings of our study should be interpreted cautiously due to the following limitations. First, the finding that MCI individuals had difficulty taking advantage of feedback under risky conditions was based entirely on behavioral manifestations. Therefore, future studies need to explore the underlying mechanisms by combining neuroimaging and electrophysiological technologies with behavioral assessments. Second, the correlation between social cognition and IGT performance in MCI was small, though significant. It remains inconclusive and needs further investigation. Third, although our study excluded individuals with apathy, it did not measure other potential decision-making factors, such as previous occupational status, cognitive reserve, and wisdom. Last but not least, the significant variation of decision-making performance in both groups might also be explained by the neuropathological heterogeneity of the study participants. Therefore, it is warranted that future studies should examine the characteristics of different subtypes of MCI. Meanwhile, the real-time recording may aid in detecting subtle changes and elucidate the potential mechanism of the decision-makingprocess.

Conclusion

In summary, the present study suggests that the individuals with mild cognitive impairment could not use feedback effectively to gain in the dice game when the rules were explicitly interpreted. In contrast, individuals with mild cognitive impairment and cognitively normal older adults perform similarly when exposed to ambiguous situations. The association between cognitive function and the decision-making process might be divergent regarding situations of risk (risky or ambiguous) and cognitive status (impaired or normal). However, further studies are warranted to elucidate the characteristics of decision-making performance in mild cognitive impairment, understand the potential psychological and neural mechanism of process hidden message in an ambiguous decision-making situation, and explore how cognitive abilities interplay with the decision-making process.

Footnotes

ACKNOWLEDGMENTS

The authors thank all research participants for the time and effort dedicated to the study. This work was supported by the National Key Research & Development Project, Ministry of Science and Technology (2017YFC1311100, 2018YFC1314200). The funding sources did not play any role in the design and conduct of the study; collection, analysis, and interpretation of data; and preparation of the manuscript, or in the decision to submit the paper for publication.

Authors’ disclosures available online ().

AVAILABILITY OF DATA AND MATERIALS

The databases are not freely available to the public because of a lack of specific approval for the ethics committee. Still, they will be available from the corresponding author on a reasonable request.

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

References

Mitchell

, Shiri-Feshki

(2008) Temporal trends in the long term risk of progression of mild cognitive impairment: A pooled analysis. J Neurol Neurosurg Psychiatry 79, 1386–1391.

Bangma

, Tucha

, De Deyn

, Koerts

(2021) How well do people living with neurodegenerative diseases manage their finances? A meta-analysis and systematic review on the capacity to make financial decisions in people living with neurodegenerative diseases. Neurosci Biobehav Rev 127, 709–739.

Okonkwo

, Griffith

, Copeland

, Belue

, Lanza

, Zamrini

, Harrell

, Brockington

, Clark

, Raman

, Marson

(2008) Medical decision-making capacity in mild cognitive impairment: A 3-year longitudinal study. Neurology 71, 1474–1480.

Parmigiani

, Del Casale

, Mandarelli

, Barchielli

, Kotzalidis

, D’Antonio

, Di Vita

, de Lena

, Ferracuti

(2021) Decisional capacity to consent to treatment and research in patients affected by mild cognitive impairment. A systematic review and meta-analysis. Int Psychogeriatrics, doi: 10.1017/S1041610220004056.

Darby

, Dickerson

(2017) Dementia, decision making, and capacity. Harv Rev Psychiatry 25, 270–278.

Hsu

, Bhatt

, Adolphs

, Tranel

, Camerer

(2005) Neuroscience: Neural systems responding to degrees of uncertainty in human decision-making. Science 310, 1680–1683.

Gleichgerrcht

, Ibáañez

, Roca

, Torralva

, Manes

(2010) Decision-making cognition in neurodegenerativediseases. Nat Rev Neurol 6, 611–623.

Ligneul

(2019) Sequential exploration in the Iowa gambling task: Validation of a new computational model in a large dataset of young and old healthy participants. PLoS Comput Biol 15, e1006989.

Boyle

, Yu

, Schneider

, Wilson

, Bennett

(2019) Scam awareness related to incident Alzheimer dementia and mild cognitive impairment: A prospective cohort study. Ann Intern Med 170, 702–709.

10.

Stewart

, Yu

, Wilson

, Bennett

, Boyle

(2019) Healthcare and financial decision making and incident adverse cognitive outcomes among older adults. J Am Geriatr Soc 67, 1590–1595.

11.

Gaubert

, Chainay

(2021) Decision-making competence in patients with Alzheimer’s disease: A review of the literature. Neuropsychol Rev 31, 267–287.

12.

Zamarian

, Weiss

, Delazer

(2011) The impact of mild cognitive impairment on decision making in two gambling tasks. J Gerontol B Psychol Sci Soc Sci 66, 23–31.

13.

Pertl

M-T

, Benke

, Zamarian

, Delazer

(2015) Decision making and ratio processing in patients with mild cognitive impairment. J Alzheimers Dis 48, 765–779.

14.

Pertl

, Benke

, Zamarian

, Delazer

(2017) Effects of healthy aging and mild cognitive impairment on a real-life decision-making task. J Alzheimers Dis 58, 1077–1087.

15.

Sun

, Xie

, Wang

, Zhang

, Tian

, Wang

, Yu

, Wang

(2020) Decision-making under ambiguity or risk in individuals with Alzheimer’s disease and mild cognitive impairment. Front Psychiatry 11, 218.

16.

Bayard

, Jacus

, Raffard

, Gely-Nargeot

(2014) Apathy and emotion-based decision-making in amnesic mild cognitive impairment and Alzheimer’s disease. Behav Neurol 2014, 231469.

17.

Bayard

, Jacus

, Raffard

, Géely-Nargeot

(2015) Conscious knowledge and decision making under ambiguity in mild cognitive impairment and Alzheimer disease. Alzheimer Dis Assoc Disord 29, 357–359.

18.

De Siqueira

ASS

, Yokomizo

, Jacob-Filho

, Yassuda

, Aprahamian

(2017) Review of decision-making in game tasks in elderly participants with Alzheimer disease and mild cognitive impairment. Dement Geriatr Cogn Disord 43, 81–88.

19.

Kloeters

, Bertoux

, O’Callaghan

, Hodges

, Hornberger

(2013) Money for nothing - Atrophy correlates of gambling decision making in behavioural variant frontotemporal dementia and Alzheimer’s disease. Neuroimage Clin 2, 263–272.

20.

Delazer

, Sinz

, Zamarian

, Benke

(2007) Decision-making with explicit and stable rules in mild Alzheimer’s disease. Neuropsychologia 45, 1632–1641.

21.

Sinz

, Zamarian

, Benke

, Wenning

, Delazer

(2008) Impact of ambiguity and risk on decision making in mild Alzheimer’s disease. Neuropsychologia 46, 2043–2055.

22.

Gupta

, Duff

, Denburg

, Cohen

, Bechara

, Tranel

(2009) Declarative memory is critical for sustained advantageous complex decision-making. Neuropsychologia 47, 1686–1693.

23.

Han

, Boyle

, James

, Yu

, Bennett

(2015) Mild cognitive impairment is associated with poorer decision-making in community-based older persons. J Am Geriatr Soc 63, 676–683.

24.

Torralva

, Dorrego

, Sabe

, Chemerinski

, Starkstein

(2000) Impairments of social cognition and decision making in Alzheimer’s disease. Int Psychogeriatr 12, 359–368.

25.

, Zhu

, Mu

, Chen

, Dong

, Cheng

, Hu

, Zhu

, Wang

(2015) Theory of mind and decision-making processes are impaired in Parkinson’s disease. Behav Brain Res 279, 226–233.

26.

Burgio

, Delazer

, Meneghello

, Pertl

, Semenza

, Zamarian

(2018) Cognitive training improves ratio processing and decision making in patients with mild cognitive impairment. J Alzheimers Dis 64, 1213–1226.

27.

Malek-Ahmadi

, Chen

, Perez

, He

, Mufson

(2018) Cognitive composite score association with Alzhei-mer’s disease plaque and tangle pathology. Alzheimers Res Ther 10, 90.

28.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

29.

Hughes

, Berg

, Danziger

, Coben

, Martin

(1982) A new clinical scale for the staging of dementia. Br J Psychiatry 140, 566–572.

30.

Folstein

, Folstein

, McHugh

(1975) Mini-Mental State: A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

31.

Yin

, Ren

, Wang

, Li

, Hou

, Liu

, Cong

, Zhang

, Wang

, Jiang

, Du

(2020) The power of the Functional Activities Questionnaire for screening dementia in rural-dwelling older adults at high-risk of cognitive impairment. Psychogeriatrics 20, 427–436.

32.

World Health Organization (2013) The ICD-10 Classification of Mental and Behavioural Disorders. Clinical descriptions and diagnostic guidelines.

33.

Mckhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

34.

Cummings

(1997) The Neuropsychiatric Inventory: Assessing psychopathology in dementia patients. Neurology 48, S10–S16.

35.

Zhou

, Zhu

, Zhang

, Wang

, Li

, Lv

, Ng

, Yu

, Wang

(2014) The influence of education on Chinese version of Montreal cognitive assessment in detecting amnesic mild cognitive impairment among older people in a Beijing rural community. ScientificWorldJournal 2014, 689456.

36.

Huang

, Wang

, Zhang

, Du

, Jia

, Wang

, Zhang

(2021) Is geriatric depression scale a valid instrument to screen depression in Chinese community-dwelling elderly? BMC Geriatr 21, 310.

37.

Wang

, Fan

, Shi

, Xiong

, Zhang

, Li

, Sun

, Guo

, Tian

, Qu

, Zhang

, Cheng

, Wu

, Han

, Tian

, Xie

, Tan

, Gao

, Luo

, Pan

, Peng

, Qin

, Tang

, Wang

, Zhang

, Zhao

, Gauthier

, Yu

(2019) Consensus statement on the neurocognitive outcomes for early detection of mild cognitive impairment and Alzheimer dementia from the Chinese Neuropsychological Normative (CN-NORM) Project. J Glob Health 9, 020320.

38.

Donohue

, Sperling

, Salmon

, Rentz

, Raman

, Thomas

, Weiner

, Aisen

(2014) The preclinical Alzheimer cognitive composite: Measuring amyloid-related decline. JAMA Neurol 71, 961–970.

39.

Sun

, Wang

, Xu

, Lv

, Li

, Wang

, Yu

(2019) Characteristics of cognitive deficit in amnestic mild cognitive impairment with subthreshold depression. J Geriatr Psychiatry Neurol 32, 344–353.

40.

Toplak

, Sorge

, Benoit

, West

, Stanovich

(2010) Decision-making and cognitive abilities: A review of associations between Iowa Gambling Task performance, executive functions, and intelligence. Clin Psychol Rev 30, 562–581.

41.

Bechara

, Damasio

, Anderson

(1994) Insensitivity to future consequences following damage to human prefrontal cortex. Cognition 50, 7–15.

42.

Song

, Xu

, Chen

, Hu

, Ge

, Xue

, Qi

, Lin

, Chen

(2021) Functional MRI-specific alterations in salience network in mild cognitive impairment: An ALE meta-analysis. Front Aging Neurosci 13, 695210.

43.

Liang

, Wang

, Yang

, Jia

, Li

(2011) Functional disconnection and compensation in mild cognitive impairment: Evidence from DLPFC connectivity using resting-state fMRI. PLoS One 6, e22153.

44.

Talwar

, Kushwaha

, Chaturvedi

, Mahajan

(2021) Systematic review of different neuroimaging correlates in mild cognitive impairment and Alzheimer’s disease. Clin Neuroradiol 31, 953–967.

45.

Glover

, Yu

, Stewart

, Wilson

, Bennett

, Boyle

(2021) The association of late life cognitive activity with healthcare and financial decision-making in community-dwelling, nondemented older adults. Am J Geriatr Psychiatry 29, 117–125.

46.

Maia

T V.

, McClelland

(2004) A reexamination of the evidence for the somatic marker hypothesis: What participants really know in the Iowa gambling task. Proc Natl Acad Sci U S A 101, 16075–16080.

47.

Denburg

, Tranel

, Bechara

(2005) The ability to decide advantageously declines prematurely in some normal older persons. Neuropsychologia 43, 1099–1106.

48.

Denburg

, Recknor

, Bechara

, Tranel

(2006) Psychophysiological anticipation of positive outcomes promotes advantageous decision-making in normal older persons. Int J Psychophysiol 61, 19–25.

49.

, Zang

, Cheung

, Chan

CCH

(2015) Importance of punishment frequency in the Iowa gambling task: An fMRI study. Brain Imaging Behav 9, 899–909.

50.

Brand

, Labudda

, Markowitsch

(2006) Neuropsychological correlates of decision-making in ambiguous and risky situations. Neural Networks 19, 1266–1276.

51.

McCullough

, Bayles

, Bouldin

(2019) Language performance of individuals at risk for mild cognitive impairment. J Speech Lang Hear Res 62, 706–722.

52.

Poudel

, Riedel

, Salo

, Flannery

, Hill-Bowen

, Eickhoff

, Laird

, Sutherland

(2020) Common and distinct brain activity associated with risky and ambiguous decision-making. Drug Alcohol Depend 209, 107884.

53.

Poletti

, Enrici

, Adenzato

(2012) Cognitive and affective Theory of Mind in neurodegenerative diseases: Neuropsychological, neuroanatomical and neurochemical levels. Neurosci Biobehav Rev 36, 2147–2164.