Table 1
Descriptive Statistics (Means and Standard Deviations) for Demographic and Clinical Measures by Group and Session.
| FULL SAMPLE | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 131 | 48 | 83 | 48 | 83 | |||||
| Age | 34.17 (10.18) | 32.19 (11.19) | 35.32 (9.46) | N/A | N/A | HC: 48 SUD +: 83 Total: 131 | t(129) = –1.7 p = 0.09 d = –0.3 | N/A | N/A |
| Sex (Male) | 114 (44%) | 22 (46%) | 35 (42%) | N/A | N/A | HC: 48 SUD +: 83 Total: 131 | χ2(1) = 0.17 p = 0.68 | N/A | N/A |
| DAST | 3.48 (3.72) | 0.12 (0.39) | 7.55 (2.17) | 0.45 (0.55) | 2.72 (3.01) | HC: 38 SUD +: 81 Total: 119 | F(1, 117) = 193.62 p < 0.001 η2 = 0.62 | F(1, 117) = 177.72 p < 0.001 η2 = 0.6 | F(1, 117) = 102.46 p < 0.001 η2 = 0.47 |
| PHQ | 3.67 (5.1) | 0.88 (1.33) | 6.84 (6.11) | 1.08 (1.81) | 3.35 (4.69) | HC: 38 SUD +: 80 Total: 118 | F(1, 116) = 30.51 p < 0.001 η2 = 0.21 | F(1, 116) = 25.15 p < 0.001 η2 = 0.18 | F(1, 116) = 16.44 p < 0.001 η2 = 0.12 |
| OASIS | 3.7 (4.35) | 1.48 (2.01) | 5.99 (4.74) | 1.39 (2.27) | 3.74 (4.49) | HC: 38 SUD +: 81 Total: 119 | F(1, 117) = 27.47 p < 0.001 η2 = 0.19 | F(1, 117) = 177.72 p < 0.001 η2 = 0.6 | F(1, 117) = 8.82 p = 0.004 η2 = 0.07 |
| WRAT | 60.42 (6.26) | 63.78 (4.61) | 58.45 (6.31) | N/A | N/A | HC: 45 SUD +: 77 Total: 122 | t(120) = 4.94 p < 0.001 d = 0.9 | N/A | N/A |
| Regular Nicotine Smoker* | 68 (26%) | 7 (15%) | 27 (33%) | N/A | N/A | HC: 46 SUD +: 47 Total: 93 | χ2(1) = 17.87 p < 0.001 | N/A | N/A |
| PROPENSITY-MATCHED | |||||||||
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 70 | 45 | 25 | 45 | 25 | |||||
| Age | 32.4 (10.37) | 32.27 (11.25) | 32.65 (8.91) | N/A | N/A | HC: 45 SUD +: 25 Total: 70 | t(68) = –0.14 p = 0.89 d = –0.04 | N/A | N/A |
| Sex (Male) | 68 (49%) | 21 (47%) | 13 (52%) | N/A | N/A | HC: 45 SUD +: 25 Total: 70 | χ2(1) = 0.18 p = 0.67 | N/A | N/A |
| DAST | 2.31 (3.4) | 0.13 (0.4) | 7.6 (2.36) | 0.45 (0.55) | 3.76 (3.37) | HC: 38 SUD +: 25 Total: 63 | F(1, 61) = 186.8 p < 0.001 η2 = 0.75 | F(1, 61) = 24.9 p < 0.001 η2 = 0.29 | F(1, 61) = 60.05 p < 0.001 η2 = 0.5 |
| PHQ | 2.74 (4) | 0.87 (1.36) | 7.68 (5.15) | 1.08 (1.81) | 3.72 (3.96) | HC: 38 SUD +: 25 Total: 63 | F(1, 61) = 63.99 p < 0.001 η2 = 0.51 | F(1, 61) = 6.59 p = 0.01 η2 = 0.1 | F(1, 61) = 16.84 p < 0.001 η2 = 0.22 |
| OASIS | 3.01 (3.8) | 1.44 (1.99) | 6.72 (4.46) | 1.39 (2.27) | 4.56 (4.33) | HC: 38 SUD +: 25 Total: 63 | F(1, 61) = 40.52 p < 0.001 η2 = 0.4 | F(1, 61) = 2.69 p = 0.11 η2 = 0.04 | F(1, 61) = 5.59 p = 0.02 η2 = 0.08 |
| WRAT | 63.29 (4.83) | 63.78 (4.61) | 62.4 (5.22) | N/A | N/A | HC: 45 SUD +: 25 Total: 70 | t(68) = 1.14 p = 0.26 d = 0.28 | N/A | N/A |
| Regular Nicotine Smoker* | 32 (23%) | 7 (16%) | 9 (36%) | N/A | N/A | HC: 44 SUD +: 17 Total: 61 | χ2(1) = 8.69 p = 0.003 | N/A | N/A |
[i] * Defined as >3650 lifetime cigarettes. DAST = Drug Abuse Screening Test. PHQ = Patient Health Questionnaire. OASIS = Overall Anxiety Severity and Impairment Scale. WRAT = Wide Range Achievement Test. Significant effects are bolded.
Table 2
Lifetime DSM-IV/DSM-5 psychiatric disorders within SUDs.
| FULL DATASET | PROPENSITY-MATCHED | |||||
|---|---|---|---|---|---|---|
| BASELINE (N = 147) | FOLLOW-UP (N = 83) | ANALYSIS | BASELINE (N = 49) | FOLLOW-UP (N = 25) | ANALYSIS | |
| Substance Use Disorders | ||||||
| Alcohol | 55 (37%) | 30 (36%) | χ2(1) = 0.04 p = 0.85 | 20 (41%) | 10 (40%) | χ2(1) = 0 p = 0.95 |
| Cannabis | 55 (37%) | 23 (28%) | χ2(1) = 2.23 p = 0.14 | 17 (35%) | 5 (20%) | χ2(1) = 1.71 p = 0.19 |
| Stimulants | 104 (71%) | 61 (73%) | χ2(1) = 0.2 p = 0.66 | 35 (71%) | 18 (72%) | χ2(1) = 0 p = 0.96 |
| Opioids | 56 (38%) | 30 (36%) | χ2(1) = 0.09 p = 0.77 | 25 (51%) | 14 (56%) | χ2(1) = 0.16 p = 0.68 |
| Sedatives | 38 (26%) | 21 (25%) | χ2(1) = 0.01 p = 0.93 | 14 (29%) | 9 (36%) | χ2(1) = 0.43 p = 0.51 |
| Hallucinogens | 5 (3%) | 3 (4%) | χ2(1) = 0.01 p = 0.93 | 2 (4%) | 2 (8%) | χ2(1) = 0.5p = 0.48 |
| 2+ Disorders | 94 (64%) | 51 (61%) | χ2(1) = 0.14 p = 0.71 | 34 (69%) | 18 (72%) | χ2(1) = 0.05 p = 0.82 |
| Alcohol Only | 9 (6%) | 6 (7%) | χ2(1) = 0.11 p = 0.74 | 3 (6%) | 2 (8%) | χ2(1) = 0.09 p = 0.76 |
| Cannabis Only | 9 (6%) | 4 (5%) | χ2(1) = 0.17 p = 0.68 | 4 (8%) | 2 (8%) | χ2(1) = 0 p = 0.98 |
| Stimulants Only | 26 (18%) | 17 (20%) | χ2(1) = 0.27 p = 0.6 | 6 (12%) | 3 (12%) | χ2(1) = 0 p = 0.98 |
| Opioids Only | 8 (5%) | 5 (6%) | χ2(1) = 0.03 p = 0.85 | 2 (4%) | 0 (0%) | χ2(1) = 1.05 p = 0.31 |
| Sedatives Only | 0 (0%) | 0 (0%) | NA | 0 (0%) | 0 (0%) | NA |
| Mood, Anxiety, Stress Disorders | ||||||
| Major Depressive | 78 (53%) | 43 (52%) | χ2(1) = 0.03 p = 0.85 | 30 (61%) | 15 (60%) | χ2(1) = 0.01 p = 0.92 |
| Generalized Anxiety | 22 (15%) | 14 (17%) | χ2(1) = 0.15 p = 0.7 | 9 (18%) | 7 (28%) | χ2(1) = 0.91 p = 0.34 |
| Social Anxiety | 19 (13%) | 11 (13%) | χ2(1) = 0.01 p = 0.94 | 8 (16%) | 3 (12%) | χ2(1) = 0.24 p = 0.62 |
| Panic | 17 (12%) | 10 (12%) | χ2(1) = 0.01 p = 0.91 | 7 (14%) | 3 (12%) | χ2(1) = 0.07 p = 0.79 |
| Post-Traumatic Stress | 23 (16%) | 14 (17%) | χ2(1) = 0.06 p = 0.81 | 10 (20%) | 5 (20%) | χ2(1) = 0 p = 0.97 |
| 2+ Disorders | 46 (31%) | 30 (36%) | χ2(1) = 0.56 p = 0.45 | 18 (37%) | 10 (40%) | χ2(1) = 0.08 p = 0.78 |
[i] Note: Stimulants = amphetamine, methamphetamine, and/or cocaine.
Figure 1
Upper left: Illustration of the three-armed bandit task interface. In each of 20 games, participants had 16 opportunities (trials) to choose between one of three options with unknown (but stable) probabilities of winning vs. not winning a point (corresponding to the appearance of a green vs. red circle above the chosen option). Throughout the task, the interface displayed the game number, trial number, total points earned, and history of wins/losses for each choice within the current game (number of green and red circles above each option; see main text for more details). Left panel: Graphical depiction of the computational (partially observable Markov decision process) model used with the task (described in the main text). The values of variables in blue circles are inferred on each trial, whereas parameter values in white circles are fixed on each trial. Here, arrows indicate dependencies between variables such that observations for each modality m (reward and observed choice) at a time t depend on choice states (st) at time t, where these relationships, , are specified by a matrix A. States depend on both previous states and the choice of action policy (π), as specified by policy-dependent transition matrices Bπ that encode p(st+1|st, π). States at t = 1 have an initial state prior specified by a vector D. Here, D = [1 0 0 0]T, such that the participant always started in an undecided ‘start’ state at the beginning of each trial. The probability of selecting an action policy depends on its expected free energy (Gπ), which in turn depends on the subjective reward value of making different observations (e.g., a win vs. loss) for the participant (in a vector C). These preferences are defined as a participant’s log-expectations over observations, . As shown in the top-right panel, the values in C are passed through a softmax (normalized exponential) function, σ(), which transforms them into a proper probability distribution, and then converted into log probabilities. Top right panel: Specifies the mathematical form of the dependencies between C, Gπ, π, and a in action selection. When there is no uncertainty about states (as is true of this task), Gπ assigns higher values to actions that are expected to simultaneously maximize information gain and reward. The first term on the right corresponds to expected information gain under approximate posterior beliefs (q). Large values for this first term indicate the expectation that beliefs about reward probabilities (A) will undergo a large change (i.e., that a lot will be learned about these probabilities) given a choice of policy, due to the states and observations it is expected to generate. The second term on the right motivates reward maximization, where a high reward value corresponds to a precise prior belief over a specific observation, . For example, if the subjective value of a win in C were cr = 4 (see bottom right panel), this would indicate a greater subjective reward (higher prior probability) than cr = 2. The policy expected to maximize the probability of a win (under the associated beliefs about states, observations, and reward probabilities) is therefore favored. Because the two terms in expected free energy are subtracted, policies associated with high expected reward and high expected information gain will be assigned a lower expected free energy. This formulation entails that information-seeking dominates when reward probabilities are uncertain, while reward-seeking dominates when uncertainty is low. A softmax function, σ(), then transforms the negative expected free energies into a probability distribution over policies, such that policies with lower expected free energies are assigned higher probabilities. When actions are subsequently sampled from the posterior distribution over policies, randomness in chosen actions is controlled by an action precision parameter (a). Bottom panel: After each observation of a win/loss, learning corresponds to updating beliefs in a Dirichlet distribution (a) over the likelihood matrix A that encodes reward probabilities. Here, columns indicate (from left to right) a starting state (pre-choice) and choices 1, 2, and 3, where the rows (from top to bottom) indicate the pre-choice (no reward) observation, observing reward, or no reward. The value of a0 – the insensitivity to information parameter – is the starting value for beliefs about reward probabilities. These beliefs always start by making up an uninformative (flat) distribution, but higher starting values (e.g., 5 vs. 0.5) effectively down-weight the information-gain term in the expected free energy – leading to an insensitivity to the need for information. The values within a (reward) are then updated based on the bottom equation, controlled by a learning rate parameter (η). For more details regarding the associated mathematics, see the main text and supplemental materials, as well as (Da Costa et al., 2020; K. J. Friston, Lin, et al., 2017; K. J. Friston, Parr, & de Vries, 2017; Smith, Friston, & Whyte, 2022). Estimated model parameters are shown in dark red.
Table 3
Computational model description.
| MODEL ELEMENT | GENERAL DESCRIPTION | MODEL SPECIFICATION |
|---|---|---|
| One vector per modality (m) of possible observations. Each vector contains entries corresponding to possible observable stimuli for that category at time t. | Possible observations for reward (modality 1):
Possible observations for choice (modality 2):
| |
| st | A vector containing entries corresponding to the probability of each possible state that could be occupied at time t. | Possible choice states:
|
| A | A matrix encoding the relationship between states and observations (one matrix per outcome modality). |
|
| a | Dirichlet priors associated with the A matrix that specify beliefs about the mapping from states to observations. Learning corresponds to updating the concentration parameters for these priors after each observation, where the magnitude of the updates is controlled by a learning rate parameter η (see Supplementary Materials and Figure 1). | Each entry for learnable reward probabilities began with a uniform concentration parameter value of magnitude a0, and was updated after each observed win or loss on the task. The learning rate η and a0 (which can be understood as a measure of sensitivity to new information; see Supplementary Materials) were fit to participant behavior. |
| B p(st+1|st,π) | A set of matrices encoding the probability of transitioning from one state to another given the choice of policy (π). Here policies simply include the choice of each bandit. | Transition probabilities were deterministic mappings based on a participant’s choices such that, for example, p(schoice 1|sstart,πoption 1) = 1, and 0 for all other transitions, and so forth for the other possible choices. |
| C p(ot) | One vector per observation modality (per time point) encoding the preference (subjective reward value) of each possible observation within that modality. This vector is passed through a softmax function and then log-transformed. | The value of observing a win was a model parameter cr reflecting subjective reward value (reward sensitivity); the value of all other observations was set to 0. The value of cr was fit to participant behavior. Crucially, higher cr values have the effect of reducing goal-directed exploration, as the probability of each choice (based on expected free energy Gπ) becomes more driven by reward than by information-seeking (see Supplementary Materials and Figure 1). |
| D p(st=1) | A vector encoding prior probabilities over states. | This encoded a probability of 1 that the participant began in the start state. |
| π | A vector encoding the probability of selecting each allowable policy (one entry per policy). The value of each policy is determined by its expected free energy (Gπ), which depends on a combination of expected reward and expected information gain. Actions at each time point are chosen based on sampling from the distribution over policies, π = σ (–G); the determinacy of action selection is modulated by an inverse temperature or action precision parameter α (see Supplementary Materials and Figure 1). | This included 3 allowable policies, corresponding to the choice of transitioning to each of the three choice states. The action precision parameter α was fit to participant behavior. |
Table 4
Nested models.
| PARAMETER: | a (ACTION PRECISION) | cr (REWARD SENSITIVITY) | η (LEARNING RATE) | a0 (INSENSITIVITY TO INFORMATION) |
|---|---|---|---|---|
| Default value if not estimated | 4 | (always estimated) | (removed from model) | 0.25 |
| Prior means during estimation* | 4 | 4 | 0.5 | 0.25 |
| Model 1 | Y | Y | N | N |
| Model 2 | Y | Y | Y | N |
| Model 3 | Y | Y | Y | Y |
| Model 4 | N | Y | Y | Y |
| Model 5 | N | Y | Y | N |
| Model 6 | N | Y | N | N |
| Model 7 | N | Y | N | Y |
| Model 8 | Y | Y | N | Y |
| Model 9** | Y | Y | Wins/Losses | Y |
| Model 10 | Y | Y | Wins/Losses | N |
[i] Y indicates that a parameter was estimated for that model; N indicates that a parameter was not estimated for that model.
* Prior variance for all parameters was set to a precise value of 2–2 in order to deter over-fitting.
** Winning model.
Table 5
Model Parameters by Group and Session (Means and Standard Deviations) as well as Results of Linear Mixed Effects Model Analyses.
| FULL SAMPLE | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 131 | 48 | 83 | 48 | 83 | |||||
| Action Precision | 2.44 (0.83) | 2.57 (0.92) | 2.2 (0.58) | 2.71 (0.95) | 2.44 (0.86) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 4.52 p = 0.04 η2 = 0.04 | F(1, 120) = 4.41 p = 0.04 η2 = 0.04 | F(1, 120) = 0.67 p = 0.41 η2 = 0.01 |
| Reward Sensitivity | 4.45 (1.49) | 4.3 (1.47) | 4.25 (1.5) | 4.5 (1.45) | 4.71 (1.49) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 0.98 p = 0.32 η2 = 0.01 | F(1, 120) = 12.74 p < 0.001 η2 = 0.1 | F(1, 120) = 1.04 p = 0.31 η2 = 0.01 |
| Learning Rate (Wins) | 0.5 (0.13) | 0.47 (0.12) | 0.5 (0.13) | 0.49 (0.13) | 0.51 (0.15) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 5.41 p = 0.02 η2 = 0.04 | F(1, 120) = 3.86 p = 0.05 η2 = 0.03 | F(1, 120) = 0.01 p = 0.93 η2 = 0 |
| Learning Rate (Losses) | 0.38 (0.15) | 0.43 (0.13) | 0.39 (0.15) | 0.39 (0.15) | 0.35 (0.16) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 6.45 p = 0.01 η2 = 0.05 | F(1, 120) = 8.68 p = 0.004 η2 = 0.07 | F(1, 120) = 0.14 p = 0.7 η2 = 0 |
| Information Insensitivity | 0.78 (0.28) | 0.72 (0.27) | 0.79 (0.29) | 0.76 (0.31) | 0.82 (0.25) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 3.63 p = 0.06 η2 = 0.03 | F(1, 120) = 1.54 p = 0.22 η2 = 0.01 | F(1, 120) = 0.27 p = 0.6 η2 = 0 |
| PROPENSITY MATCHED | |||||||||
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 70 | 45 | 25 | 45 | 25 | |||||
| Action Precision | 2.51 (0.89) | 2.59 (0.94) | 2.12 (0.58) | 2.67 (0.94) | 2.51 (0.9) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 2.43 p = 0.12 η2 = 0.04 | F(1, 68) = 2.86 p = 0.1 η2 = 0.04 | F(1, 68) = 1.63 p = 0.21 η2 = 0.02 |
| Reward Sensitivity | 4.4 (1.51) | 4.23 (1.47) | 4.17 (1.63) | 4.51 (1.49) | 4.72 (1.52) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 0.02 p = 0.88 η2 = 0 | F(1, 68) = 4.84 p = 0.03 η2 = 0.07 | F(1, 68) = 0.59 p = 0.45 η2 = 0.01 |
| Learning Rate (Wins) | 0.49 (0.13) | 0.46 (0.12) | 0.53 (0.1) | 0.49 (0.13) | 0.51 (0.16) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 2.56 p = 0.11 η2 = 0.04 | F(1, 68) = 0.31 p = 0.58 η2 = 0 | F(1, 68) = 1.48 p = 0.23 η2 = 0.02 |
| Learning Rate (Losses) | 0.39 (0.15) | 0.43 (0.13) | 0.35 (0.17) | 0.39 (0.14) | 0.34 (0.18) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 4.32 p = 0.04 η2 = 0.06 | F(1, 68) = 1.33 p = 0.25 η2 = 0.02 | F(1, 68) = 0.65 p = 0.42 η2 = 0.01 |
| Information Insensitivity | 0.77 (0.29) | 0.73 (0.27) | 0.82 (0.33) | 0.75 (0.28) | 0.86 (0.28) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 3.23 p = 0.08 η2 = 0.05 | F(1, 68) = 0.35 p = 0.56 η2 = 0.01 | F(1, 68) = 0.07 p = 0.8 η2 = 0 |
[i] * Analyses are reported using results from LMEs accounting for age, sex, and premorbid IQ (WRAT). Significant effects are bolded.
Figure 2
Left: Results of parametric empirical Bayes (PEB) analyses, showing the posterior means and variances for group difference estimates in the full and propensity-matched samples in models accounting for age, sex, and premorbid IQ. These Bayesian group comparisons confirm the differences in learning rates for losses seen at baseline. There was also a main effect of time on this learning rate, but no significant interactions between group and time, indicating the group effects were stable. No other parameters showed main effects of group. See main text for further results of these analyses. Learning rate values are in logit-space. Right: Spaghetti plots showing individual changes from baseline to follow-up, as well as group means and standard errors, for learning rate for losses in the full and matched samples. HCs = healthy controls, SUDs = substance use disorders.
Table 6
Intra-class correlations between baseline and 1-year follow-up (full sample).
| GROUP | ICC(3, 1) | p | |
|---|---|---|---|
| Total wins | All | .15 | .05 |
| HCs | .27 | .03 | |
| SUDs | .08 | .23 | |
| a (action precision) | All | .32 | <.001 |
| HCs | .45 | <.001 | |
| SUDs | .15 | .09 | |
| cr (reward sensitivity) | All | .54 | <.001 |
| HCs | .48 | <.001 | |
| SUDs | .58 | <.001 | |
| ηwin (learning rate for wins) | All | .35 | <.001 |
| HCs | .28 | .03 | |
| SUDs | .37 | <.001 | |
| ηloss (learning rate for losses) | All | .43 | <.001 |
| HCs | .35 | .007 | |
| SUDs | .45 | <.001 | |
| a0 (insensitivity to information) | All | .25 | .002 |
| HCs | .24 | .05 | |
| SUDs | .25 | .01 |
Figure 3
Correlations between computational parameters at baseline and 1-year follow-up.
Figure 4
Top: Negative correlation in stimulant users (full sample) between pre-to-post changes in action precision and pre-to-post changes in symptom severity (DAST). Bottom: Illustration of individual pre-to-post changes in DAST scores and action precision (as well as group mean and SE). As can be seen, DAST scores tend to decrease and action precision tends to increase, but with notable individual differences in each. DAST change scores account for what could already be predicted based on age, sex, and premorbid IQ. However, we note that this correlation did not survive correction for multiple comparisons and will need to be replicated in future work.
Figure 5
Predictive relationships in stimulant users (full sample) between baseline model parameters and symptom severity at 1-year follow-up, after accounting for what could already be predicted based on age, sex, and premorbid IQ. The p-values shown here are uncorrected, but the relationships with learning rates and insensitivity to information survive correction for 6 comparisons (i.e., one per SUD group tested; corrected threshold of p < .0083). No relationship survives a more conservative correction for 30 comparisons (i.e., accounting for 5 parameters tested in each SUD group; corrected threshold of p < .0017).
Table 7
Model-Free Task Measures by Group and Session (Means and Standard Deviations).
| FULL SAMPLE | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 131 | 48 | 83 | 48 | 83 | |||||
| Wins | 181.1 (12.71) | 182.83 (12.19) | 178.75 (12.79) | 182.27 (12.14) | 181.78 (13.12) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 1.97 p = 0.16 η2 = 0.02 | F(1, 120) = 1.33 p = 0.25 η2 = 0.01 | F(1, 120) = 0.74 p = 0.39 η2 = 0.01 |
| Mean Reaction Time | 0.56 (0.25) | 0.62 (0.24) | 0.61 (0.27) | 0.53 (0.23) | 0.5 (0.22) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 3.46 p = 0.07 η2 = 0.03 | F(1, 120) = 29.56 p < 0.001 η2 = 0.2 | F(1, 120) = 0.6 p = 0.44 η2 = 0.01 |
| Win/Stay | 134.28 (33.91) | 133.5 (32.91) | 131.63 (36.63) | 131.21 (31.49) | 139.17 (33.03) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 0.27 p = 0.6 η2 = 0 | F(1, 120) = 2.61 p = 0.11 η2 = 0.02 | F(1, 120) = 1.99 p = 0.16 η2 = 0.02 |
| Win/Shift | 35.08 (28.17) | 37.88 (28.11) | 35.46 (29.6) | 39.06 (25.85) | 30.8 (27.94) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 1.42 p = 0.24 η2 = 0.01 | F(1, 120) = 2.15 p = 0.15 η2 = 0.02 | F(1, 120) = 1.24 p = 0.27 η2 = 0.01 |
| Lose/Stay | 47.11 (29.45) | 39.33 (25.51) | 46.42 (30.11) | 45.98 (24.45) | 52.96 (32.66) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 7.21 p = 0.008 η2 = 0.06 | F(1, 120) = 9.16 p = 0.003 η2 = 0.07 | F(1, 120) = 0.06 p = 0.8 η2 = 0 |
| Lose/Shift | 83.52 (30.99) | 89.29 (27.39) | 86.49 (32.07) | 83.75 (27.07) | 77.07 (33.27) | HC: 45 SUD +: 77 Total: 122 | F(1, 117) = 4.08 p = 0.05 η2 = 0.03 | F(1, 120) = 12.7 p < 0.001 η2 = 0.1 | F(1, 120) = 0.79 p = 0.38 η2 = 0.01 |
| PROPENSITY-MATCHED | |||||||||
| TOTAL | BASELINE | FOLLOW-UP | USABLE DATA (N) | EFFECT OF CLINICAL STATUS | EFFECT OF SESSION | EFFECT OF CLINICAL STATUS/SESSION INTERACTION | |||
| HCS | SUDS | HCS | SUDS | ||||||
| 70 | 45 | 25 | 45 | 25 | |||||
| Wins | 180.76 (12.56) | 182.4 (12.46) | 177.92 (11.15) | 182.44 (11.95) | 177.64 (14.65) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 3.44 p = 0.07 η2 = 0.05 | F(1, 68) = 0 p = 0.97 η2 = 0 | F(1, 68) = 0.01 p = 0.93 η2 = 0 |
| Mean Reaction Time | 0.55 (0.23) | 0.62 (0.25) | 0.57 (0.21) | 0.53 (0.23) | 0.47 (0.17) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 1.87 p = 0.18 η2 = 0.03 | F(1, 68) = 14.39 p < 0.001 η2 = 0.17 | F(1, 68) = 0.04 p = 0.85 η2 = 0 |
| Win/Stay | 132.44 (32.63) | 132.96 (33.22) | 129.64 (38.41) | 132.24 (32.13) | 134.68 (27.58) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 0 p = 0.97 η2 = 0 | F(1, 68) = 0.09 p = 0.76 η2 = 0 | F(1, 68) = 0.39 p = 0.54 η2 = 0.01 |
| Win/Shift | 36.78 (27.93) | 38.11 (28.05) | 37.04 (33.09) | 38.09 (26.38) | 31.76 (25.9) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 0.46 p = 0.5 η2 = 0.01 | F(1, 68) = 0.28 p = 0.6 η2 = 0 | F(1, 68) = 0.49 p = 0.49 η2 = 0.01 |
| Lose/Stay | 46.19 (29.26) | 38.31 (25.88) | 51 (30.01) | 45.22 (24.7) | 57.32 (37.94) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 4.56 p = 0.04 η2 = 0.07 | F(1, 68) = 3.97 p = 0.05 η2 = 0.06 | F(1, 68) = 0.01 p = 0.93 η2 = 0 |
| Lose/Shift | 84.59 (29.69) | 90.62 (27.71) | 82.32 (32.22) | 84.44 (27.78) | 76.24 (33.13) | HC: 45 SUD +: 25 Total: 70 | F(1, 65) = 2.04 p = 0.16 η2 = 0.03 | F(1, 68) = 3.13 p = 0.08 η2 = 0.04 | F(1, 68) = 0 p = 0.99 η2 = 0 |
[i] * Analyses are reported using results from LMEs accounting for age, sex, and premorbid IQ (WRAT). Significant effects are bolded.