Delay discounting and reward sensitivity in a 2x2 study of bipolar disorder and alcohol dependence

Abstract

Background and Aims:

Separate studies have shown increased delay discounting in people with bipolar disorder (BD) and people with alcohol dependence (AD) relative to people without mental health problems. Compared delay discounting in people in people with no mental health problems, AD, BD and AD plus BD. Associations of delay discounting with self-reported impulsivity and reward sensitivity were also assessed

Design:

Two-by-two factorial comparative observational design.

Setting:

Data were collected at baseline diagnostic visits as part of a neuroimaging study at a medical university in South Carolina, USA.

Participants:

Twenty-two BD+AD, 33 BD, 28 AD, and 27 people without mental health problems.

Measurements:

Diagnostic and clinician-rated symptom measures, self-report questionnaires, and a computerized delay discounting task were administered. Two-by-two general linear univariate models were tested to examine between-group differences on discounting rates, and bivariate correlations and hierarchical regression analyses were performed to examine associations between discounting rates and self-reported reward sensitivity and impulsivity.

Findings:

There was a significant main effect of AD (p = 0.006, η² = 0.068). The main effect of BD and the BD x AD interaction term were non-significant (p ≥ 0.293, η² ≤ 0.010). Reward sensitivity and trait impulsivity were not significantly associated with discounting rates after adjustment for the other (p ≥ 0.089).

Conclusions:

People with alcohol dependence appear to have higher delay discounting while previously found associations between bipolar disorder and delay discounting may be secondary to alcohol use disorder.

Introduction

Bipolar disorder (BD) is the psychiatric disorder most commonly associated with alcohol dependence (AD) (1). Prominent shared neurobehavioral characteristics of BD and AD, including elevated impulsivity and reward sensitivity, are proposed to contribute high rates of co-occurrence between BD and AD (BD+AD) (2–4). Empirical investigation of these clinically relevant variables may inform multiple areas of inquiry and help shape future assessment and treatment of BD, AD, and BD+AD.

Delay (or temporal) discounting refers to the decrease in value of a reward as a function of its delay to receipt (5). The degree to which an individual discounts reward, their choice impulsivity, is indicated by their preference for smaller sooner rewards over larger delayed rewards in a two-choice task (e.g., choosing $99 today rather than $100 in 1 day) (6–8). After making several reward-delay choices, the individual’s “indifference point” for each delay (the subjective value at which the immediate and delayed reward is equally preferred) is calculated. From this data, an individual’s discounting curve, represented by kappa (k), can be produced which provides the rate at which the subjective value of reward decreases as a function of its time to receipt (7, 9, 10). Delay discounting is relevant to various psychiatric disorders and is therefore well studied (11, 12), albeit it surprisingly less so in BD.

Heavy-drinking and AD individuals discount reward more steeply than light, social-drinking controls (13–17). Less is known about delay discounting in BD. In fact, only two known studies have been conducted. Similar to findings in AD, a study by (18) found adults with BD to more steeply discount rewards than healthy controls. In a related study (19), adolescents with bipolar-spectrum disorders did not significantly differ from healthy adolescents on discounting rates, though a trend-level increase was associated with bipolar-spectrum disorder status. Most clinical participants met criteria for bipolar disorder not otherwise specified, implying a lack of symptom severity/duration which may explain discrepant results from (18).

Prior research has shown self-report and behavioral impulsivity to be questionably associated (14, 20, 21). In fact, most studies have found the two to weakly correlate at best (15, 17, 18, 22–24). A related construct that may be worthy associative candidate of choice impulsivity is reward sensitivity (25, 26). Reward sensitivity mediates approach behavior, motivation for appetitive stimuli, and positive affect (27, 28). Heightened sensitivity to reward is influential to the onset and course of BD, AD, and BD+AD illness (4, 29–31). Reward sensitivity therefore may serve as an important discounting correlate with potential for improving the neurobehavioral characterization of these patient populations.

Against this background, the present investigation was the first delay discounting study of BD+AD, BD, and AD individuals. Aims were two-fold:

• To compare delay discounting rates between individuals with BD+AD, BD, AD, and healthy controls. BD and AD were hypothesized to have significant main effects (7, 13, 18); however, whether these effects would significantly interact in BD+AD was unclear (3). Regardless, all clinical groups were expected to discount reward more steeply than healthy controls.

• To examine dimensional associations between self-reported impulsivity and reward sensitivity, and delay discounting rates across the full sample. This aim was exploratory.

Materials and methods

Participants and procedures

Participants.

One-hundred ten individuals who met DSM-IV-TR criteria for BD I or II and current AD (criteria met over preceding 6 months) (BD+AD, n = 22), BD I or II alone (BD, n = 33), current AD alone (AD, n = 28), or healthy controls (HC, n = 27) who did not meet criteria for any psychiatric disorder were included. Eight additional participants were recruited but excluded due to missing delay discounting data (n = 6) or for being extreme outliers on delay discounting rates (n = 2; k values ≥ 3^rd quartile + 1.5 x the interquartile range). Study completers were recruited through community advertisements (42.6%), from clinical settings on campus including research referrals (29.1%) and inpatient (10.2%) and outpatient referrals (9.8%), and by word-of mouth (8.3%).

Inclusion/exclusion criteria were predetermined by Prisciandaro et al.’s (2017) 2x2 proton magnetic resonance spectroscopy and functional neuroimaging study of BD and AD. A primary aim of that study was to better describe characteristics of BD+AD relative to BD and AD alone. Inclusion criteria stated that BD+AD and AD participants could have co-occurring drug abuse or dependence and/or anxiety disorder(s). BD participants, in turn, could have co-occurring anxiety disorder but not meet criteria for substance abuse or dependence. These determinations were made in service of the neuroimaging aims of the parent study and to increase study feasibility and generalizability. Exclusionary criteria included: serious medical illness or history of head injury, psychotic disorder, recurrent major depressive disorder, past-month post-traumatic stress disorder, obsessive-compulsive disorder or eating disorder (assessed with the Structured Clinical Interview for DSM-IV Axis I Disorders [SCID-IV;(32)]); current use of benzodiazepines or antidipsotropics (e.g., naltrexone, topirimate, acamprosate, disulfiram); past-month electroconvulsive therapy, history of delirium tremens or > 1 lifetime alcohol withdrawal seizure, acute alcohol withdrawal (> 10 on the Clinical Institute Withdrawal of Alcohol Scale [CIWA-Ar; (33)]), and past-month daily drug use. BD+AD and AD participants were required to exceed healthy weekly drinking levels (> 7/week for females, > 14/week for males) over the past month (or month preceding abstinence). BD and HC participants were to report ≤ healthy weekly drinking levels. Seventy-two hours of drug and alcohol abstinence was required at baseline. Urine drug screens and ethyl glucuronide (EtG) samples were collected to corroborate self-report. Cigarette smoking status was defined as an average daily consumption of 10 or more cigarettes (34).

The study was conducted at the Medical University of South Carolina (MUSC) using a protocol approved by the institutional review board. Written informed consent was obtained from every participant and participation was entirely voluntary. Present study data were collected at baseline diagnostic visits as part of Prisciandaro et al. (3).

Measures

The Structured Clinical Interview for DSM-IV-TR Axis I disorders (SCID-IV) (32) was conducted to determine the presence or absence of DSM-IV-TR psychiatric diagnoses. The clinician-administered Young Mania Rating Scale (YMRS) (35) and Montgomery-Asberg Depression Rating Scale (MADRS) (36) were used to measure past-week (hypo)manic and depressive symptoms, respectively. The YMRS consists of 11 items with a maximum score of 60 whereas the MADRS consists of 10 items with a maximum score of 60. Higher scores on each measure indicate greater symptom levels. The self-reported Barratt Impulsiveness Scale (BIS-11) consists of three subscales measuring distinct facets of impulsivity: Attentional (8 items), Motor (11 items), and Non-Planning (11 items). Maximum scores for Motor and Non-Planning subscales are 44 whereas the maximum score for the Attentional subscale is 32. Reward sensitivity was measured with the Sensitivity to Reward scale of the self-reported Sensitivity to Punishment and Sensitivity to Reward Questionnaire (SPSR-Q) (27) which consists of 24 items with a maximum score of 24. For each of these self-report measures, higher scores indicate greater levels of impulsivity and reward sensitivity, respectively.

Delay discounting.

The delay discounting task (37) was administered via computer using E-Prime 2.0 software (38). Instructions were provided by research staff, participants demonstrated understanding, and then completed the task in private. Choices were made for 70 unique pairs of hypothetical monetary reward-delay options, with a 15s maximum for responding with 2s inter-stimulus intervals. Sample items include: “Would you rather have $100 in 1 day or $99 today?” and “Would you rather have $80 today or $100 in 1 month?” Hypothetical rewards were chosen for practical purposes (16). Responses were recorded via dominant-hand keyboard presses of “1” or “2,” made for the selection of the left or right choice presentations shown on screen. Values included $10, $20, $30, $40, $60, $80, $99, and $100 with time delays of today, 1 day, 1 week, 1 month, 3 months, 6 months, and 1 year.

Data analytic strategy

Discounting rates (k) were calculated using the standard hyperbolic equation, V = A/(1 + kD), where “V” is the subjective value of the reward, “A” is the reward amount, and “D” is the delay to reward. The hyperbolic, rather than exponential, discounting function was used because abundant research shows this function best fits human choice behavior across various populations, substance-dependent and not alike (8, 11, 15, 40, 41). Indifference points were determined by taking the average of the last larger reward and first delayed reward chosen, and k values were calculated using nonlinear regression (42, 43). Consistent with the literature, k values were highly skewed and non-normal so a base-10 logarithm transformation was employed to make data suitable for parametric statistics (41, 44). The model fit of the hyperbolic function to the observed data was evaluated by examining the percentage of variance (R²) explained in the actual subjective value of each delayed reward (45). The hyperbolic function accounts for approximately 85% of variance in human choice behavior (46).

Prior to performing main study analyses, variable data were visualized and tested for skewness and kurtosis to ensure the appropriateness of parametric statistical approaches. Bivariate Pearson correlations and hierarchical linear regression analyses were used to examine associations between k and other variables of interest. This approach along with chi-square tests of independence, independent t-tests, and One-Way Analyses of Variance (ANOVAs) with post-hoc LSD tests were performed to evaluate potential covariates for primary study analyses. In the event of insufficient cell sizes (< 5 participants) for chi-square tests, Fisher’s Exact test statistic was used (47). 2x2 factorial general linear models (GLMs) with dichotomous BD and AD variables as fixed factors, from which a BD+AD interaction term was formed, were performed to examine group differences on k. Effect sizes were reported in partial eta-squared (η²). The conventional p < 0.05 threshold was employed in evaluating statistical significance of results. Any non-significant group effects in 2x2 GLM analyses were subsequently assessed with Bayes factor calculations to determine the likelihood that effects were indeed null, at the population level, given the observed data (48).

A model building approach was taken in that analyses were performed with and without covariate adjustment since sample-dependent variables may explain some outcome variance. Potential covariates, submitted to statistical evaluation, were identified a priori in an objective manner (49). Tobacco and other drug dependencies, for instance, associate with steeper delay discounting making these variables potential confounders (7, 45, 50). Analyses were performed using a Microsoft Excel macro and SPSS Version 25.0 (39).

Results

Hyperbolic model fit

Prior to evaluating potential covariates and conducting primary analyses, the fit of the hyperbolic function to the data was assessed to ensure adequate model fit, thereby allowing for meaningful interpretation of results. The hyperbolic function provided excellent fit to full sample data (46), median R² = 0.90. Within-group median R² values were similarly excellent: 0.89 for BD+AD, 0.90 for BD, 0.91 for AD and 0.91 for HCs.

Sample characteristics

See Table 1 for participant characteristics by diagnostic group and results from group comparisons. As shown, BD+AD and BD participants were similar on BD-related variables (i.e., BD subtype, number of past-year [hypo]manic and depressive episodes) whereas BD+AD and AD participants were similar on AD-related variables (i.e., alcohol dependence severity, heavy drinking days [%] over past 90 days). Though excluded from the table, across the full sample the most prevalent co-occurring drug dependence diagnoses were cannabis dependence and cocaine dependence (44.7% and 39.5% of those reporting drug dependence, respectively). The prevalence of lifetime anxiety disorder was greatest for Social Phobia (32.1% of all anxiety disorder diagnoses) followed by PTSD (28.3%), GAD (15.1%), and Specific Phobia (9.4%).

Table 1.

Participant characteristics and group comparison results

	Participant group				p
	BD+AD1 (n = 22)	BD alone2 (n = 33)	AD alone3 (n = 28)	HC4 (n = 27)
Age (in years)	37.00(12.62)	36.94(11.89)	41.82(11.41)	38.15(10.46)	0.355
Sex (% female)	45.5	54.5	32.1	48.1	0.365
Race (% Caucasian)	86.4	75.8	64.3	81.5	0.308
Education (% completed 4-yr college)	31.8	53.6	51.5	37.0	0.304
Current smoker (%)	36.42	9.11,3,4	35.72	33.32	0.057
Drug dependence (%)	81.82,4	0.01,3	75.02, 4	0.01,3	0.000
Anxiety disorder (%)	59.13,4	46.73,4	21.41,2, 4	0.01−3	0.000
BD diagnosis (age)	26.82(8.78)3,4	26.27(9.34)3,4	0.01,2	0.01,2	0.829
BD subtype (% Type-I)	72.73,4	60.63,4	0.01,2	0.01,2	0.000
YMRS	3.45(3.61)3,4	2.87(3.80)4	1.82(1.83)1,4	0.22(0.51)1−3	0.000
MADRS	11.45(8.98)3,4	9.61(6.71)3,4	5.65(5.54)1,2,4	1.19(1.80)1−3	0.000
# past-year (hypo)manic episodes	4.07(5.19)3,4	2.11(2.98)3,4	0.01,2	0.01,2	0.000
# past-year depressive episodes	1.86(1.70)3,4	1.64(2.22)3,4	0.01,2	0.01,2	0.000
ADS	18.48(7.32)2,4	2.25(2.74)1,3	17.00(8.14)2,4	0.76(1.64)1,3	0.000
% Heavy-drinking days (past 90)	45.92,4	54.61,3	2.52,4	3.01,3	0.000
Past-month number of daily drinks	7.83(6.31)2,4	0.28(0.62)1,3	7.23(4.72)2,4	0.29(0.33)1,3	0.000
% active in AA/NA	36.42,4	0.01,3	46.42,4	0.01,3	0.000
BIS Attentional	19.54(4.15)3,4	18.00(5.28)4	15.35(4.40)1,4	11.52(2.93)1−3	0.000
BIS Motor	26.45(4.97)4	24.06(5.59)4	23.43(4.70)4	18.93(3.25)1−3	0.000
BIS Non-planning	30.86(5.19)3,4	26.59(6.08)4	24.75(6.59)1	21.33(5.92)1,2	0.000
SR	14.28(4.63)2,4	9.72(4.86)1	12.22(4.53)4	6.88(3.66)1,3	0.000
Medication (%)
Lithium	13.63,4	21.23,4	0.01,2	0.01,2	0.002
Antipsychotic	40.93,4	54.83,4	0.01,2	0.01,2	0.000
Anticonvulsant	63.62−4	35.51,3,4	0.01,2	0.01,2	0.000
Antidepressant	54.53,4	32.33,4	0.01,2	0.01,2	0.000

Data are mean(standard deviation) unless otherwise noted. Abbreviations: AD, alcohol dependence; BD, bipolar disorder; YMRS, Young Mania Rating Scale; MADRS, Montgomery-Asberg Depression Rating Scale; YMRS, Young Mania Rating Scale; ADS, Alcohol Dependence Scale; NIAAA-defined “heavy drinking” days are > 4 drinks/day for men or > 3 drinks/day for women; AA, alcoholics anonymous; NA, narcotics anonymous; BIS Attentional, Motor, and Non-planning, Barratt Impulsiveness Scale-11 Subscales; SR, Sensitivity to Reward Scale of the Sensitivity to Punishment and Sensitivity to Reward Questionnaire. Smoking status indicates smoking 10 or more cigarettes per day. Drug Dependence represents past-year drug dependence whereas Anxiety disorder represents lifetime anxiety disorder diagnoses. P < 0.10 omnibus F (continuous variables) and χ² (categorical variables) tests were followed up with pairwise tests. Within-cell superscript numbers denote which participant groups a given group was significantly different from (p < 0.05) in these follow-up comparisons.

Bivariate associations between study variables

See Table 2 for Pearson bivariate correlations between main study variables across participants. Manic symptoms and reward sensitivity were the only variables to significantly associate with logk. Drug dependence remained non-significantly associated with logk even when examined in BD+AD and AD participants, specifically (p-values ≥ 0.319). Additionally, there were no significant independent associations between any medication class (i.e., Lithium) and logk when examined in the BD+AD and BD participants, specifically (p-values ≥ 0.195).

Table 2.

Bivariate associations between main study variables

	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. Logk	-
2. Age	0.13 (0.193) n = 110	-
3. Sex	−0.13 (0.164) n = 110	−0.15 (0.128) n = 110	-
4. Race	−0.11 (0.245) n = 110	−0.15 (0.126) n = 110	−0.05 (0.598) n = 110	-
5. Education	−0.18 (0.055) n = 110	0.09 (0.341) n = 110	0.03 (0.728) n = 110	0.24 (0.11) n = 110	-
6. Smoking	0.15 (0.130) n = 104	0.24 (0.016) n = 104	−0.19 (0.052) n = 104	−0.12 (0.245) n = 104	−0.27 (0.006) n = 104	-
7. Drug dependence	0.14 (0.161) n = 108	−0.11 (0.248) n = 108	−0.17 (0.082) n = 108	−0.08 (0.455) n = 108	−0.13 (0.182) n = 108	0.26 (0.007) n = 103	-
8. Anxiety disorder	0.00 (0.987) n = 109	−0.16 (0.108) n = 109	−0.03 (0.729) n = 109	0.09 (0.365) n = 109	−0.14 (0.141) n = 109	0.01 (0.957) n = 104	0.30 (0.002) n = 108	-
9. Manic symptoms	0.24 (0.013) n = 108	0.11 (0.244) n = 108	0.09 (0.359) n = 108	−0.14 (0.157) n = 108	−0.08 (0.414) n = 109	−0.09 (0.376) n = 102	0.07 (0.451) n = 106	0.21 (0.031) n = 107	-
10. Depressive symptoms	0.06 (0.555) n = 110	−0.09 (0.329) n = 110	0.22 (0.023) n = 110	0.07 (0.488) n = 110	0.06 (0.545) n = 110	−0.26 (0.007) n = 104	0.04 (0.663) n = 108	0.44 (0.000) n = 109	0.29 (0.003) n = 108	-
11. Attentional impulsivity	0.06 (0.571) n = 109	−0.19 (0.047) n = 109	0.17 (0.084) n = 109	0.24 (0.010) n = 109	−0.03 (0.732) n = 109	−0.05 (0.638) n = 103	0.14 (0.138) n = 107	0.46 (0.000) n = 108	0.29 (0.003) n = 107	0.63 (0.000) n = 109	-
12. Motor impulsivity	0.15 (0.115) n = 109	−0.11 (0.276) n = 109	−0.02 (0.876) n = 109	0.25 (0.010) n = 109	−0.11 (0.278) n = 109	0.05 (0.594) n = 103	0.18 (0.059) n = 107	0.30 (0.002) n = 108	0.28 (0.004) n = 107	0.41 (0.000) n = 109	0.74 (0.000) n = 109	-
13. Non-planning impulsivity	0.18 (0.069) n = 109	−0.12 (0.198) n = 109	0.08 (0.433) n = 109	0.22 (0.022) n = 109	−0.12 (0.226) n = 109	−0.05 (0.644) n = 103	0.23 (0.017) n = 107	0.46 (0.000) n = 108	0.26 (0.006) n = 107	0.53 (0.000) n = 109	0.73 (0.000) n = 109	0.61 (0.000) n = 109	-
14. Reward sensitivity	0.21 (0.035) n = 106	−0.13 (0.189) n = 106	−0.24 (0.013) n = 106	0.08 (0.429) n = 106	−0.06 (0.521) n = 106	−0.01 (0.938) n = 101	0.42 (0.000) n = 104	0.26 (0.007) n = 105	0.07 (0.455) n = 104	0.30 (0.002) n = 106	0.43 (0.000) n = 105	0.47 (0.000) n = 105	0.40 (0.000) n = 105	−

Data are Pearson correlation(p-value). Logk = discounting rate, Smoking = current consumption of 10+ cigarettes per day, Drug dependence = presence of any drug dependency, Anxiety disorder = presence of any anxiety disorder, Manic symptoms = YMRS total score, Depressive symptoms = MADRS total score, Attentional impulsivity = BIS-11 Attentional impulsivity subscale, Motor impulsivity = BIS-11 Motor impulsivity subscale, Non-planning impulsivity = BIS-11 Non-planning impulsivity subscale, Reward sensitivity = SR total score from SRSP-Q.

Delay discounting rates between groups

2x2 GLM univariate analyses were performed to compare delay discounting rate between participant groups. BD and AD were fixed factors, from which a BD x AD interaction term formed. Initial covariates selected for adjustment included smoking status, drug dependence, and anxiety disorder (given the significant difference between AD and HCs despite it not being a prerequisite for group assignment). Taking a model building approach, non-significant covariates were discarded until a final, parsimonious model was identified. See Table 3 for model output. This resulted in a basic 2x2 model, which demonstrated a significant main effect of AD, a non-significant main effect of BD, and a non-significant BD x AD interaction. See Figure 1 for a visual depiction of findings.

Table 3.

2x2 GLM examining group effects on delay discounting rate (logk)

Preliminary model
Factor	F	p	η2
BD	2.33	0.131	0.024
AD	6.27	0.014	0.061
BD x AD	0.02	0.833	0.000
Smoking status	2.82	0.096	0.029
Drug dependence	1.52	0.221	0.016
Anxiety disorder(s)	0.42	0.521	0.004
Parsimonious model excluding non-significant covariates
BD	1.12	0.293	0.010
AD	7.75	0.006	0.068
BD x AD	0.006	0.940	0.000

BD = Bipolar Disorder; AD = Alcohol Dependence, and BD x AD = Co-occurring Bipolar Disorder and Alcohol Dependence. Smoking status = ≥ 10 cigarettes consumed/day. Drug dependence and Anxiety disorder = presence/absence of disorder(s).

Figure 1.

Delay discounting rate by diagnostic group

Data are mean logk with ± 1 standard error bars; higher values indicate steeper discounting rate.

Bayes factors (BF) were calculated for BD and BD+AD terms to evaluate the degree of confidence that factor effects are indeed non-significant given the observed data (48). Results showed that a true null effect of BD (BF = 0.10) was 10 times more likely than an undetected, significant effect. More tentative results supported the probability of a true, non-significant BD x AD interaction effect (BF = 0.63).

Reward sensitivity and trait impulsivity associations with delay discounting rate

See Table 4 for final (step 2) hierarchical regression results when logk was regressed on self-reported reward sensitivity and impulsivity across full sample data. The regression model was non-significant, F (4, 104) = 2.08, p = 0.089, R² = 0.08, as were all predictor variables. Results remained unchanged when also modeling drug dependence and smoking status. However, the model was significant, F (5, 102) = 3.26, p = 0.009, R² = 0.14, when including manic symptoms as a step 1 covariate (p = 0.005, 95% CI = 0.03, 0.18) and manic symptoms remained the only significant predictor variable at step 2 (p–values ≥ 0.088). Variance inflation factor (median = 2.47) and tolerance indices (median = 0.44) showed multicollinearity was not a threat among predictors. BF calculation showed that truly non-significant effects were likely given the observed data (BF = 0.013).

Table 4.

Model regressing self-reported reward sensitivity and trait impulsivity on delay discounting rates

	Unstandardized coefficient (B)	Standard error	Standardized beta coefficient (β)	t	p	95% confidence interval (B)
Constant	−2.78	0.49	-	−5.64	0.000	−3.76, −1.80
Attentional impulsivity	0.04	0.02	0.17	1.51	0.135	−0.01, 0.08
Motor impulsivity	−0.06	0.03	−0.28	−1.72	0.089	−0.13, 0.01
Non-planning impulsivity	0.03	0.03	0.14	0.97	0.336	−0.03, 0.09
Reward sensitivity	0.04	0.02	0.22	1.52	0.131	−0.01, 0.08

N = 104. Attentional impulsivity = BIS-11 Attentional Impulsivity, Motor impulsivity = BIS-11 Motor Impulsivity subscale, Non-planning impulsivity = BIS-11 Non-planning Impulsivity subscale, Reward sensitivity = SR scale from SPSR-Q.

Discussion

This delay discounting study was the first to examine BD+AD and the first to directly compare BD and AD, respectively. As hypothesized, AD had a significant main effect on discounting rates such that BD+AD and AD participants discounted rewards more steeply than HCs. However, contrary to hypotheses, BD and HC participants had similar discounting rates. The non-significant BD x AD interaction effect is consistent with prior self-report findings by our research group showing the two diagnoses to be non-multiplicative (3, 4). Across full sample data, neither self-reported reward sensitivity nor impulsivity significantly associated with discounting rates after statistical adjustment for one another.

Steeper delay discounting associating with AD is consistent with prior alcohol research and the greater addiction literature (11, 13–17). Null BD findings deviate from those of Ahn et al. (18) who showed steeper discounting among BD individuals. This discrepancy may be partially due to distinctive mood symptom presentations between study samples. Present inclusion/exclusion criteria stated a strict (low) (hypo)manic symptom cut-off for BD(±AD) group assignment, resulting in borderline euthymic samples (51). The BD sample in (18), in turn, reported symptom levels consistent with an emergent (hypo)manic switch in mood state (52). This difference could have been critical because hypo(manic) symptoms have been shown to inversely associate with temporal reward-related decision-making (19, 25). Another important sample characteristic consideration for the present study regards the HC sample of whom 33% were smokers, a factor known to associate with steeper discounting (45). Very few BD participants, however, were smokers (9%). Perhaps elevated smoking among HCs, paired with a euthymic mood state in BD participants, effective “washed out” any BD effect(s) thereby contributing to null findings. Then again, statistical adjustment for smoking status left findings unaltered. Howeover, (18) did not report on smoking status making the matter unclear when cross-evaluating findings. Collective BD findings, though, are highly tentative until further research on the subject is published. Given the presumptively strong connection between BD and delay discounting, the lack of extant papers on the topic is surprising and may speak to the proverbial “file drawer” phenomenon (53).

In the context of BD+AD, steeper delay discounting appears driven by effects of AD rather than BD. This role could emerge through the enhanced self-control failure associated with AD whereby individuals abandon sobriety for the immediately reinforcing effects of alcohol despite the expense of losing the greater delayed reward(s) of continued abstinence (11). The competing neurobehavioral decision systems theory (CNDS) (46, 54, 55) is an empirically-supported dual-systems model that aims to provide explanation for the self-control failure commonly observed in addiction (56). The CNDS posits two interdependent brain systems, one “impulsive” and once “executive,” exert relative control during decision-making. Healthy functioning occurs with regulatory balance while imbalance increases potential for pathological behavior (57). If functional neuroimaging were applied to BD+AD delay discounting as it has in AD (58), it is unclear if the impulsive system would be hyperactive and/or the executive system would be hypoactive, or some other possibility. What is important is that if unique between-group differences are demonstrated, with (or without) distinct associations with clinical outcomes, then pharmacological interventions may advance via greater specificity for target(s) of treatment in BD+AD and/or AD.

Non-significant associations between self-reported dimensions of impulsivity and delay discounting (choice) impulsivity is consistent with the majority of prior research (14, 15, 17, 20–24). Many of these authors have cited the non-association likely reflects the fact that distinct dimensions of impulsivity are measured by these different methods. Part of the discrepancy may arise, in part, because impulsivity is measured through implicit (self-report) versus explicit (decision-making) means (40). The fact that reward sensitivity was the only variable among the few to have significant independent associations with discounting rates implies it may be promising for further evaluation. Future research may demonstrate potentially unique associations between the two variables in different patient populations. Moreover, the two may associate as a function of time delay to reward. For instance, there may be stronger associations when specific indifference points are considered (i.e., perhaps for earlier delays rather than later ones).

This study is not without limitations. The delay discounting task used hypothetical, rather than actual, monetary reward; however, research shows comparable results regardless (44, 59–61). Consistent with prior studies (10, 18), the BIS-11 was selected to measure trait impulsivity in the present study. However, alternative measures such as the Eysenck Impulsivity Inventory (62) and the Multidimensional Personality Questionnaire Control (vs. Impulsivity) scale (63) have been used by other studies (45, 64) and were also worthy of consideration. BD(±AD) participants were not treatment-naïve to psychiatric medication(s) so it is unclear exactly whether findings generalize to patients with emergent, yet to be treated, BD(±AD). Non-significant independent associations between all classes of mood stabilizers and delay discounting outcomes hopefully quell this concern. Statistical power to detect BD or BD x AD effects could have been limited by sample sizes though bayes factors were calculated to substantiate null findings. Finally, the present study sample was ethno-racially homogeneous which could limit generalizability.

Limitations notwithstanding, the present study makes a rather unique contribution to the literature by examining delay discounting in co-occurring psychopathology. This strength of this study design lies not only in the high rates of BD and AD co-occurrence but also the dearth of research conducted with this difficult-to-treat patient population (1). As there are also strong associations between other psychiatric disorders and substance use disorder(s) (65–67), a similar approach may be taken to tease apart potentially unique effects of each condition with the ultimate goal of increasing understanding and improving clinical outcomes.