In this study, 53 subjects including 50 males and 3 females were assigned into two groups of control (n = 20) and individuals who use opioids (n = 33), and were selected based on the beta power of 95% and alpha of 0.05. The age of the participants was between 18 and 60 years old, and their level of education was between primary school and master’s degree (Supplementary Table 1). All subjects were screened for neurological conditions or medical history, and none of them had major psychiatric comorbidities or acute physical disabilities, and any intellectual disability was also considered as exclusion criteria. All procedures performed were in accordance with the ethical criteria of the APA, institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All participants entered the study with informed consent, and the study was approved by the Ethics Committee of the Iran University of Medical Sciences, No. IR.IUMS.FMD.REC.1400.108.

The individuals who use opioids were all selected based on the diagnosis of two experienced psychiatrists and according to the criteria in the Diagnostic and Statistical Manual of Mental disorders (DSM-V). All individuals who were using opioids were Persian native speakers and all had a history of opioid substance use such as smoking opium and heroin for at least 6 months. The group that use opioids was not in treatment and did not receive any treatments while admitted. They were hospitalized for 1-7 days when they were examined. Also, following the examination of our two psychiatrists, participants and controls were not diagnosed for any major mental disorders or psychiatric comorbidities, and none of them had obvious physical disabilities.

After recording demographic information such as age, gender, level of education and marital status, all participants underwent cognitive assessment using the n-back test, the WCST, and the SDA test. All persons who used opioids were in a steady state without any withdrawal signs when they completed the tasks.

The n-back task is one of the most classical and well-established cognitive paradigms for studying working memory (WM). WM is defined as a cognitive system of limited-capacity that provides temporary storage space as well as required information for cognitive functions such as learning, reasoning and language comprehension (19). The n-back task was originally introduced by Kirchner as a visuo-spatial test of four load factors (“0-back” to “3-back”) (20), and then by Mackworth as a visual letter task with up to six load factors that present letters or pictures as stimulus sequences (21). Basically, n-back task engages multiple processes, such as selection, decision making, suppression and interference separation (22). For each item in the sequence, the participant must decide whether the current stimulus matches the one displayed “n” trials ago (23). Thus, the subject not only needs the storage and continuous information updating in WM, but also requires interference resolution (24). In fact, the participant requires to monitor a series of stimuli and then to respond once the stimulus (i.e., letters, numbers or pictures) is similar to that of the previous n trials, where n is a pre-specified integer as 1, 2, or 3. Responses such as wrong response, no response and correct response as well as the reaction time (ms) were selected as four subscales of n-back task in this study.

Developed by David Grant and Esta Berg, the WCST is a neuropsychological test that is mainly used to measure executive function and higher-level cognitive abilities such as perseverance, abstract thinking, cognitive flexibility, and set shifting (25, 26). The WCST consists of four stimulus cards and 64 response cards and there are various geometric shapes in different dimensions (colors, forms and numbers). The participants are expected to identify the specific sorting rule and accurately match every response card with one of four stimulus cards through the feedback based on a rule (25). The sorting rule that the participants identify through a process of trial and error is the dimension that each card should be correctly matched. For instance, a response card with three red stars can be matched according to color (red), form (star), or number (three). Following each response, the subject receives feedback (i.e., “correct” or “incorrect”) that is employed to establish the correct sorting rule. Normally, the sorting rule changes without previous warning after ten correct responses in a row which is referred as completing a category and the subject should start the task again to establish the new sorting rule for the next category. Different subscales that were studied in this task include number of categories completed, perseverative errors, other types of errors, correct or wrong responses, number of total tries and tries to complete the first level, total time (s), conceptual level responses, and failure to continue on a specific pattern. The WCST ends when either all of the six categories are completed or 128 trials are done.

Here, the SDA test which is a continuous performance task (CPT), is divided into two modules. In the selective attention module, participants need to press a special key on the keyboard at maximum speed if they see a predetermined item, and to restrain their response if they see other stimuli (27). In the divided attention module, participants are requested to press specific keys if they see one or both of the predetermined items in their prespecified locations, and if they see other stimuli or if they see the predetermined items in the location other than their prespecified locations, they need to restrain their response (28).

The statistical indices of centrality and dispersion of the distribution, including mean and standard deviation were used to describe the demographic characteristics as well as the performance results of each group in the cognitive tests. In order to compare the nominal demographic variables between the two groups, the Fisher’s exact test was applied.

The multivariate generalized linear model (MGLM) was used to compare the dependent variables between the two groups. This model provides a regression analysis and analysis of variance for multiple dependent variables by one or more factor variables or covariates.

In order to find possible functional interactions between subscales, the correlations between each two subscales were calculated by the non-parametric Spearman test in the two groups (Supplementary Tables 2, 3). To reconstruct the cognitive interactive network, the correlation tables were loaded in the Cytoscape, version 3.8.2. Cytoscape is an open source platform for visualizing interactive cognitive networks and biological pathways in an integrative manner (29). Functional interactive networks were reconstructed in such a way that each node represents one of the evaluated subscales and each edge represents a statistically significant correlation.

In order to analyze the networks topologically and to determine the size of each node, the betweenness centrality was calculated for each node, using “igraph” package in R ver. 3.4.0 (30). Betweenness centrality, C_B, is an invariant of graph that indicates the number of times a node acts as a bridge along the shortest path between two other nodes. In other words, for a vertex v ∈ V(G):

where σ_st is the total number of shortest paths from node s to node t and σ_st(v) is the number of those paths that pass through v (31). Size of each node in the networks were adjusted to display its centrality magnitude. The larger the size of a node, the more central role it plays in transmitting information between more pairs of nodes. Based on edge betweenness, thickness of edges is adopted to represent the difference in betweenness centralities and to better differentiate the most central values.

Furthermore, closeness centrality (CC) is shown as assigned colors to each of the nodes and was used as an index of importance of a vertex within a given complex network that measures how close a vertex is to all other vertices in the graph as an index of average node distance. Nodes whose color shifted to the red spectrum have a higher CC value, meaning that they are closer to other nodes in terms of network topography. However, nodes whose color changed to the blue spectrum have a lower CC value, meaning that they are farther in relation to other nodes. Thus, the centrality of a node’s closeness indicates how far that node is on average from other associated nodes and in fact, how much they are involved in effectively directing connections in a network. The CC is calculated through the following formula:

where d(y,x)is the distance between nodes xand y, and Nis the number of nodes. The node properties including mean C_B (Supplementary Figure 2) and CC (Supplementary Figure 3) for each of the subscales are estimated separately for each group. Thickness of each edge was considered proportional to their betweenness. The red spectrum indicates positive correlations between nodes and the blue spectrum shows negative correlations. In order to find the most important links represented as edges, C_Bof each link was multiplied by its correlation coefficient and thus its weight was calculated (Supplementary Table 5).

Fisher’s exact test indicated that there is no significant difference between the two groups in terms of demographic variables including education level (p = 0.549), female to male ratio (p = 0.338), and marital status (p = 0.059; Supplementary Table 1). The mean age of the individuals who use opioids and controls was 29.94 ± 9.19 and 36.05 ± 7.75, respectively, and following the U Mann-Whitney test, there was a statistically significant difference between the two groups [z = –2.55, p = 0.011]. Therefore, age was considered as a covariate in other analyzes.

The MGLM analysis showed that the two groups differed significantly in a number of subscales including the number of tries in the WCST [F(1, 53) = 7.452, p-value = 0.009, ηp2 = 0.0171], and divided attention omission errors [F(1, 53) = 11.270, p-value = 0.002, ηp2 = 0.184], and divided attention correct responses [F(1, 53) = 7.902, p-value = 0.007, ηp2 = 0.170] in the SDA test (Figure 1, Supplementary Figure 1, and Supplementary Table 4). Furthermore, remarkably more omission errors in divided attention module during the SDA task show a potentially diminished divided attention for the group that use opioids compared to the control group. This is also corroborated by notably more correct responses in the same task and the same module performed by controls.

Following the psychological test score analysis for different task subscales and different cognitive functions, interactive network of those cognitive domains was reconstructed for both groups to illustrate the most central links and cognitive domains with significant contribution to changes in cognitive state of the individuals who use opioids compared to controls. In the reconstructed cognitive network for the subjects with opioid use disorder (Figure 2A), the reaction time (C_B = 0.39, CC = 0.58) and the number of commission errors (C_B = 0.37, CC = 0.56) in the divided attention module of the SDA test show the highest betweenness centrality (Supplementary Figure 2) and closeness centrality (Supplementary Figure 3), meaning that these two subscales are the most central for the group that use opioids. These data from the centrality measures of the cognitive network support the results from MGLM analysis implying that the persons who use opioids had a significant increase in reaction time or a much lower speed during tasks and more errors indicative of a potential loss of divided attention. Following the calculation of weight of the edges for each of the two linking nodes through multiplication of the C_B measures and correlation coefficients of every two linked subscales, the most important positive link with the largest weight was between the number of tries in the WCST and the reaction time in the divided attention module of the SDA test (43.87). This highly significant positive link between the number of tries and the reaction time justifies the more attempts in performing a certain module of the task in relation to the higher reaction time in SDA and potential loss of attention in the individuals who use opioids. Further, the most important negative link with lowest weight was found to be between the number of commission errors and the reaction time of the divided attention module in SDA test (-58.03). This largely significant negative link between the number of errors and the reaction time also justifies the more errors in performing a certain module of the task in relation to the higher reaction time in SDA and potential loss of attention in the persons who use opioids.

On the other hand, in the reconstructed cognitive network for the control subjects (Figure 2B), the number of correct responses (C_B = 0.18, CC = 0.64) had the highest betweenness centrality (Supplementary Figure 2), and the number of commission errors (C_B = 0.14, CC = 0.65) indicated the highest closeness centrality (Supplementary Figure 3) in the divided attention module of the SDA task, meaning that these two subscales are the most central for the controls. These results from the centrality measures of the cognitive network also corroborate the data from MGLM analysis implying that the controls had significant correct responses indicative of a better divided attention. Following the estimation of weight of the edges for each of the two linking nodes for every two linked subscales, the most important positive link with the largest weight was between the reaction times of the selective and the divided attention modules in the SDA test (27.27). This largely significant positive link between the reaction times of the attention module justifies the importance of reaction time in both selective and divided modules in maintaining attention during the SDA task. Further, the most important negative link with lowest weight was found to be between the number of tries to complete the first level in the WCST and the reaction time in the divided attention module of the SDA test (-18.43). This highly significant negative link between the number of tries in WCST and the reaction time in SDA also justifies the fewer attempts in performing a certain module of a task such as the first level of the WCST in relation to the lower reaction time in the divided attention module of the SDA test which indicates intact divided attention skills in the control group.