Six-month buprenorphine-naloxone treatment is associated with neurocognitive function improvement in opioid dependence.

Background and Aim: The number of longitudinal studies on cognitive functions in patients on buprenorphine-based agonist treatment is limited. Our objective was to assess the change in neurocognitive functions over the first 6 months of buprenorphine-naloxone (BNX) treatment for opioid dependence (OD) and compare cognitive functions on BNX and controls.
Methods: We selected 60 patients with OD aged 18 to 55 years and 20 sex-matched controls; and excluded patients with other substance dependence, human immunodeficiency virus (HIV), head injury, epilepsy, and severe mental illness. We assessed patients thrice: at baseline, 3, and 6 months and Controls once. Cognitive tests included Wisconsin card sorting test (WCST), Iowa gambling task (IGT), trail making tests A and B (TMT-A and B), verbal and visual N-back test (NBT), and standard progressive matrices (SPM). We measured with-in group effect size with Cohen's D (d).
Results: A total of 24 participants completed at least one follow-up; 17 completed both follow-up assessments. All participants were men. At baseline, the patients performed worse than healthy controls in IGT, TMT-A, and B, and visual and verbal NBT. At 3 months, the performance of visual NBT improved significantly (d = 1.2 for NBT1; 1.3 for NBT2). At 6 months, additional performance improvements were seen in WCST ("perseverative error" d = 1.2), IGT ("net total score" d = 1.2), TMT-A (d = 1.1), and verbal NBT ("omission error" d = 1.7). Except for visual-NBT, results did not differ between patients and controls at both follow-ups.
Conclusion: Cognitive flexibility, decision making, attention, working memory, and psychomotor speed showed improvements over 6 months of a stable dose of BNX. Copyright:

INTRODUCTION

Worldwide opioid dependence (OD) contributes substantially to the global burden of disease and disability. The Global Burden of Diseases, Injuries, and Risk Factors Study 2016 estimated that 3.6 million years of life lost was contributed by OD.[1] The significant causes of mortality included opioid overdose, suicide, hepatitis C, and HIV.[2] The United States has declared the ongoing opioid crisis as a public health emergency.[3] Buprenorphine (with or without naloxone combination) is the standard of care for treating OD.[4] Buprenorphine and methadone are the two most commonly used agonist treatments.[5] Research has consistently shown the effect of buprenorphine in reducing illicit opioid use, risks of overdose, acquisition, and transmission of hepatitis C and HIV.[678910]

Previous studies suggested adverse neurocognitive effects of medications for OD. Cross-sectional studies comparing the cognitive effects of methadone and buprenorphine with healthy controls (HCs) showed significantly impaired cognitive functions in patients on agonist medications.[11121314] The buprenorphine group performed worse in working memory, vigilance, attention, cognitive flexibility, and psychomotor speed.[121314] However, cognitive functions were better in the buprenorphine group than in methadone.[1213] Cross-sectional studies could not inform: (a) the stability of the impaired cognitive performance in the buprenorphine group and (b) whether the better neurocognitive functions in the buprenorphine-treated patients are relatively permanent effect or reflects a transient opioid switching effect. Only prospective studies could examine these questions.

The number of longitudinal studies on cognitive functions in patients on buprenorphine-based agonist treatment is limited. Rapelli et al. (2011) tested cognitive functions 2 months, 6 to 9 months, and 12 to 17 months after the initiation of buprenorphine.[15] After the first 2 months of initiation, the results showed that the working and verbal memory functions were significantly worse in the buprenorphine group than in the HC. However, the cognitive functions improved between the second and third assessments. The limitations of this study were: sample size only 14, no baseline assessment of cognitive functions, and only specific domains of memory and attention assessed.

We conducted a 6-month prospective study in patients with OD, started on buprenorphine-naloxone (BNX), using a wide range of neurocognitive function tests at three-time points—at baseline, 3 months, and 6 months. We hypothesized (a) patients on BNX would perform significantly better in the neurocognitive tests on subsequent assessments and (b) although the BNX group may show poorer performance than controls at the intake, the follow-up assessments may not show a significant difference in the cognitive functions between the groups. The objective of the study was to determine (a) the change in the neurocognitive functions over the first 6 months of BNX treatment and (b) to compare the cognitive functions of patients on BNX and HCs.

METHODS

Participants

The study recruited patients with OD, initiated on the BNX to treat OUD in a specialized addiction treatment clinic in India. The sampling was purposive. We selected patients of both sex and the age range 18 to 55 years. A qualified psychiatrist diagnosed OD that was confirmed by using Mini Neuropsychiatric International Interview (6.0).[16] We excluded patients with other substance use (except tobacco and cannabis), HIV infection, head injury, epilepsy, schizophrenia spectrum, and major affective disorders. These comorbidities may influence cognitive function and act as confounders or effect modifiers. We also recruited 20 age and sex-matched HCs who were not using drugs, alcohol, or tobacco. Controls were recruited from the accompanying family and friends. The clinical psychologist who did the psychological assessments was unaware of the “group status” during the baseline evaluation. None of the neuropsychological assessments were part of our routine clinical care. Written informed consent was obtained from all the participants before any assessment. The study was approved by Institute Ethics Committee (IEC-09/2016-516).

Sample size calculation

The study required sample size of 17 (number of pairs) to achieve a power of 80% and a level of significance of 5% (two-sided) to detect an effect size of 0.75 between pairs. The effect size was derived from the study by Rapelli et al.[15] The study found that the buprenorphine group had a 0.75 effect size of the difference of verbal working memory between two time points. From our previous clinical experience, we assumed a drop-out of 60% to 70% in the first 6 months. Therefore, the sample, adjusted for the drop-out, recruited for the baseline assessment should have ranged between 44 and 56. We stopped the recruitment as soon as the desired numbers of paired observations were completed.

Assessments

Each patient completed neurocognitive assessments at three time points: baseline, at 3 months, and at 6 months of BNX treatment. The baseline assessment was done after at least 7 (±3) days of initiation of BNX and abstinence from the illicit opioids. The interval was used to manage the withdrawal symptoms and stabilize the dose of buprenorphine, as we did not want the participants to be tested during the active opioid withdrawal. The second and the third assessments were done 3 months (±2 weeks) and 6 months (±2 weeks) after the initiation of BNX treatment. The Clinical Opioid Withdrawal Scale scores were less than five during all the assessments.[17] We carried out urine chromatographic immunoassays (CIA) using a six-drug panel (morphine, amphetamine, phencyclidine, cannabinoids, benzodiazepines, and barbiturates) and for buprenorphine and tramadol. The CIA was done to detect any recent use of illicit drugs. We excluded patients with a history of taking illicit opioids within the last three days of assessments or with a positive urine screen for any of these drugs (except for buprenorphine).

A clinical psychologist administered the following neurocognitive tests. We performed all tests in one sitting. The duration for administration was between 45 and 60 minutes.

Standard Progressive Matrices (SPM)

SPM assesses fluid intelligence that comprises inductive and deductive reasoning and is unlikely to be affected by the years of formal education.[18]

The Wisconsin Card Sorting Test (WCST)

WCST determines working memory, executive function, and cognitive flexibility. We used the 64-cards computerized version of the test. Total errors and perseverative errors are markers of cognitive flexibility, whereas non-perseverative error measures avoiding distractions. Executive function and abstract reasoning are assessed by the conceptual level response.[19]

The Iowa Gambling Task (IGT)

IGT assesses decision-making, impulsivity, and risk-taking behavior. There are four decks of cards—the first two are disadvantageous decks and the rest are advantageous decks. We used two parameters from the IGT: the net score and the total money earned.[20]

Trail Making Tests (TMT)

We applied both the TMT-A and TMT-B. TMTs assess attention, visual search and scanning, set-shifting, psychomotor speed, and flexibility. The “time taken to complete” was used for our study.[21]

N-Back Test (NBT)

We used the paper version of the verbal and visual N-Back tests to tap the verbal and visual working memory, respectively. Besides that, NBT measures impulsivity and response inhibition through commission errors. The accuracy of response in the NBT also suggests executive control.[22]

Statistical analysis

We performed the statistical analysis with the Statistical Package for Social Sciences, 14th version.[23] The descriptive clinical and demographic variables were expressed by frequencies and percentages for categorical data (e.g., marital status and income) and means and standard deviations for continuous data (e.g., age and years of education). The group of patients retained on BNX after 6 months (analyzed group) and the group dropped out from treatment (not-analyzed group) were compared by unpaired t-test. The significance level was corrected for multiple comparisons by Bonferroni correction. Totally 16 comparisons were made, and the significance level was adjusted at P <.003. The mean scores of various cognitive tests performed at three time points were compared with General Linear Model (GLM)-repeated measures. Post-hoc Tukey’s test was used for conducting the pairwise comparisons. We expressed the effect sizes as partial eta squared and Cohen’s D, resepectively. A total of 22 comparisons were performed for each pairwise comparison; hence, the P was adjusted to <.002. The means between HC and the patient group were compared by GLM, univariate. The full factorial model was used. Age and years of education were used as covariates. Levene’s test of equality tested the homogeneity of variance of age across the groups. The Bonferroni test adjusted the confidence intervals. The effect size was calculated with Cohen’s D.

RESULTS

Between April 2017 and March 2018, a total of 60 patients were enrolled in the study. Among these 60 patients, 24 completed at least one follow-up, and 17 completed follow-ups at 3 and 6 months. All participants were men.

Sample characteristics

We described clinical and demographic characteristics of the sample at baseline (T0), 3-month (T1), and 6-month follow-up (T2) in Table 1. Please note the sample sizes were 60, 24, and 17 at the three assessments.

Table 1

Clinical and demographic characteristics of the cases and controls

Parameter	Control (n=20) [n (%)/mean (SD)]	Cases (T0) (n=60) [n (%)/mean (SD)]	Cases (T1) (n=24) [n (%)/mean (SD)]	Case (T2) (n=17) [n (%)/mean (SD)]	Control vs T0			Control vs T1			Control vs T2
					t/χ2/F	df	P	t/χ2/F	df	P	t/χ2/F	df	P
Age (Years)	32 (8.2)	25.8 (5.7)	27.5 (6.9)	28.6 (7.4)	−3.7	78	<.001	−1.9	42	.054	−1.3	35	.20
Mean education (Years)	13.9 (2)	11.3 (2.4)	11.6 (2.3)	11.7 (1.9)	−4.4	78	<.004	−3.4	42	.001	−3.3	35	.002
Marital status
Not-married	8 (40)	45 (75)	14 (58.3)	9 (52.9)	8.2	1	.004	1.4	1	.23	.62	1	.43
Married	12 (60)	15 (25)	10 (41.7)	8 (47.1)
Occupation
Employed	20 (100)	41 (68.3)	11 (45.8)	10 (58.8)	28	1	<.001	12.2	1	<.001	.002	1	.048
Unemployed	0 (0)	19 (31.7)	13 (54.2)	7 (41.2)
Family income
Up to 10000	3 (15)	33 (55)	11 (45.8)	6 (35.3)	9.6	1	.006	4.8	1	.05	.25	1	.15
≥10000	17 (85)	27 (45)	13 (54.2)	11 (64.7)
Religion
Hindu	17 (85)	30 (50)	10 (41.7)	7 (41.2)	7.6	2	.02	8.9	2	.01	.03	2	.02
Sikh	3 (15)	28 (46.7)	12 (50)	8 (47.1)
Others	0 (0)	2 (3.3)	2 (8.3)	2 (11.7)
Locality
Urban	15 (75)	36 (60)	12 (50)	6 (35.3)	1.5	1	.23	2.9	1	.09	5.9	1	.02
Rural	5 (5)	24 (40)	12 (50)	11 (64.7)
Family type
Nuclear	10 (50)	35 (55)	12 (50)	8 (47.1)	.42	1	.5	0	1	1	.03	1	.85
Extended/joint	10 (50)	25 (45)	12 (50)	9 (52.9)
Age of first use of opioids (years)	NA	19.1 (3.6)	19 (2.8)	18.7 (2.3)	------------------------------NA-------------------------------
Duration of use (months)	NA	82.3 (61.2)	103.8 (69.9)	117.5 (73.4)
Duration of dependence (months)	NA	74.2 (57.4)	98.2 (66.4)	110.7 (69.1)
Mean buprenorphine dose (mg)	NA	5.7 (2.7)	5.3 (2.3)	5.6 (2.6)
Type of opioid
Heroin	NA	57 (95)	22 (91.7)	15 (88.2)
Others	NA	3 (5)	2 (8.3)	2 (11.8)
Route of administration
Intravenous	NA	55 (91.7)	22 (91.7)	16 (94.1)
Others	NA	5 (8.3)	2 (8.3)	1 (5.9)
Baseline total COWS score	NA	4.7 (5.1)	5 (5.3)	4 (5.1)
Baseline total SODQ score	NA	41.4 (10.3)	40.2 (9.8)	41 (9.8)

NA- Not applicable

The patients and controls were compared using independent sample t-tests. In sum, the controls had a higher mean year of education and income than the patient group (at all three-time points).

We used a GLM to compare performance across various cognitive function tests among the patient groups (T0, T1, and T2 pairwise comparison) and the patient and control groups; after Bonferroni correction, the significance level was. 002 Table 1.

Pairwise comparison of neurocognitive functions of patients on BNX at three-time points (T0, T1, and T2)

SPM: The mean IQ on successive follow-ups increased, but without statistically significant difference either on pairwise comparison between T0, T1; T1, T2, and T0, T2 or on within group comparison at three time points.

WCST: The mean number of total correct responses and conceptual level response was lowest at T0 and highest in the T2. The mean number of total errors, perseverative error, and non-perseverative error was highest at T0 and lowest in the T2. The mean number of perseverative errors was significantly lower at T2 than the T0 (P =0.002), with a large effect size (1.2). The pairwise differences in none of the other parameters reached statistical significance. The within group difference at three time points too did not differ.

IGT: Overall, there was a significant within group difference at three time points in both parameters of IGT. The net total score and total money earned were lowest in T0 and highest at T2. The pairwise comparison showed a statistically significant increase in both total scores (P <.001, effect size 1.4) and total money earned (P =0.002, effect size 1) at T2 (than T0) with a large effect size.

TMTs: Overall, the performance in both TMTs differed significantly at three time points. The mean time taken to complete trail-making tests (both A and B) was highest at T0 and lowest at T2. On pairwise comparison, a statistically significant reduction in time was observed between T0 and T2 with a large effect size for both trail-making A (P =0.001, effect size 1.1) and trail-making B (P =0.002, effect size 1) tests.

Verbal working memory 1 backtest: The mean number of omission, commission, and total errors of commission and omission were highest at T0 and lowest at T2, but on the pairwise comparison, the changes in none of the parameters were statistically significant. Overll, there was no within group difference in any of these parameters.

Verbal working memory 2 backtest: The mean number of omission, commission, and total errors of omission and commission were highest at the baseline assessment (T0) and lowest at 6 months follow up (T2). There was a significant reduction in errors of omission at T2 than T0 (P <.001), with a large effect size (1.7). Within group difference at three time points, too was statistically significant.

Visual working memory 1 backtest: The mean number of omission, commission, and total errors of omission and commission were highest at T0 and lowest at T2. On pairwise comparison, a statistically significant reduction in errors of commission (P <.001, effect size 1.6) and total errors of commission and omission (P <.001, effect size 1.2) was observed at T1 than at T0. There were significant reductions in the errors of commission (P<.001, effect size 1.4) and those of omission and commission (P =0.001, effect size 1.1) at T2 than T0. Overall, there was a significant within group difference in the total error score.

Visual working memory 2 backtest: The mean number of errors was highest at T0 and lowest at T2. On pairwise comparison, a statistically significant reduction in errors of omission (P <.001, effect size 1.5), commission (P =0.002, effect size 1.1), and total errors of commission and omission (P <.001, effect size 1.3) was observed at T1, compared to T0. We observed similar results between T0 and T1 time points. Overall, all parameters in this test differed significantly within group, at three time points. Table 2.

Table 2

Comparison of cognitive functions at baseline and two other time points

Parameter	T0 [mean (SD)] (n=17)	T1 [mean (SD)] (n=17)	T2 [mean (SD)] (n=17)	T0 vs T1 vs T2			T0 vs T1			T1 vs T2			T0 vs T2
				F	Df	P	MD	Std error	P (ES)	MD	Std error	P (ES)	MD	Std error	P (ES)
Standard progressive Matrix IQ	96.7 (11.7)	102.4 (10)	104.7 (15.1)	4.8	2	.02	−5.7	2.2	.06	−2.3	2.3	1	−8	2.5	.02
Wisconsin Card Sorting Test
Total correct	38.5 (10.9)	46.1 (9.9)	47.7 (8.8)	5.5	2	.02	7.6	2.7	.03	1.5	2.7	1	−9.1	2.7	.01
Total error	25.5 (10.9)	18.8 (8.9)	16.3 (8.8)	5	2	.02	6.6	2.4	.04	2.5	2	.66	9.2	2.8	.01
Perseverative error	14.1 (6.2)	9.3 (4.6)	7.8 (5.1)	7.5	2	.01	4.8	1.4	.01	1.5	1.1	.55	6.2	1.5	.002 (1.2)
Non-perseverative error	11.4 (6.0)	9.5 (5.6)	8.5 (5)	1.5	2	.25	1.9	1.5	.66	1.1	1.4	1	2.9	1.8	.35
Conceptual level responses	31 (14.7)	39.4 (12.5)	42.6 (13.1)	4.3	2	.03	8.4	3.4	.08	3.3	2.9	.83	11.6	3.8	.02
Iowa Gambling Task
Net total score	3.3 (24.8)	31.8 (34)	47.2 (29.3)	16.4	2	<.001 (0.69)*	28.5	7.8	.01	15.4	6.3	.08	43.9	7.9	<.001 (1.4)
Total money earned	−712.6 (1327.5)	480.3 (1696)	998.9 (1664.7)	10.5	2	.001 (.58)	1192.9	349.7	.01	518.6	341.7	.45	1711.5	404.7	.002 (1.1)
Trail A (time in sec)	89.2 (29.6)	67.2 (29.7)	58.3 (19.8)	11.9	2	.001 (.61)*	21.9	8.1	.04	8.9	4	.11	30.9	7.1	.001 (1.1)
Trail B (time in sec)	127.8 (53.7)	97.8 (46)	83.2 (22.8)	11.8	2	.001 (.61)*	30.1	14.2	.15	14.6	7.6	.22	44.6	10.7	.002 (1)
Verbal Working Memory 1 back
Omission	.4 (.7)	.2 (.4)	.1 (.2)	5.9	2	.01	.2	.2	.9	.1	.1	.2	.4	.2	.09
Commission	.2 (.8)	.1 (.2)	0 (0)	1	2	.33	.2	.2	1	.1	.1	1	.2	.2	.6
Errors of omission and commissions	.7 (1.3)	.4 (.5)	.2 (.4)	1.7	2	.21	.4	.4	1	.2	.1	.6	.5	.3	.4
Verbal Working Memory 2 back
Omission	2.4 (1)	1.4 (.9)	.5 (.7)	14.6	2	<.001 (.66)*	1	.3	.03	.9	.3	.01	1.9	.3	<.001 (1.7)
Commission	.6 (.8)	.1 (.3)	0 (0)	8.7	2	.003	.5	.2	.1	.1	.1	.50	.6	.2	.01
Errors of omission and commissions	3.1 (1.5)	1.3 (1)	1.1 (2.2)	8,5	2	.003	1.8	.5	.004	.2	.6	1	2	.7	.03
Visual Working Memory 1 back
Omissions	2.1 (1.4)	1.6 (.8)	1.4 (.71)	5.7	2	.01	.5	.3	.3	.3	.2	.3	.8	.3	.03
Commissions	1.8 (.9)	.6 (.9)	.5 (.5)	2.2	2	.15	1.1	.2	<.001 (1.6)	.2	.2	1.00	1.3	.2	<.001 (1.4)
Errors of omission and commissions	3.9 (1.7)	2.3 (1.2)	2.1 (1.0)	10.5	2	<.001 (.58)*	1.6	.3	<.001 (1.2)	.2	.2	1.00	1.8	.4	.001 (1.1)
Visual Working Memory 2 back
Omission	3.1 (1)	1.4 (.87)	2.2 (.6)	13.0	2	.001 (.63)*	1.8	.3	<.001 (1.5)	.8	.2	.015	1	.2	.001 (1)
Commission	2.4 (1.2)	1.1 (1.0)	.4 (.6)	21.9	2	<.001 (.74)*	1.3	.3	.002 (1.1)	.7	.24	.03	2	.3	<.001 (2.2)
Errors of omission and commission	5.5 (1.8)	3.3 (1)	2.5 (1)	51.7	2	<.001 (.87)*	2.2	.4	<.001 (1.3)	.8	.3	.03	3	.3	<.001 (2.3)

The corrected level of significance after Bonferroni correction = .002; Effect size estimation: Cohen’s d; 0.2-0.5 small effect size; 0.5-0.8 medium effect size; >0.8 large effect size; *Effect size partial eta squared

Comparison of neurocognitive functions of patients and the control group at three-time points (T0, T1, and T2)

SPM: The mean SPM IQ was higher in the HC group than in the patient group (at all times). However, the group differences were statistically non-significant.

WCST: HC group had more correct responses and conceptual level responses and fewer total, perseverative, and non-perseverative errors than the patient group. However, the differences were not statistically significant.

IGT: The net score and money earned in the control group were higher than those of T0, T1, and T2. In comparison to the HCs, the net total score (P <.001, effect size. 52) and total money earned (P <.001, effect size. 26) were significantly smaller in the T0 group. This significant difference did not persist on follow-ups.

TMTs: In comparison to HCs, patients in the T0 group took significantly longer time to complete trail-making tests A (P <.001, effect size. 17) and B (P <.001, effect size. 15), with moderate effect size. These significant differences did not persist on follow-ups.

Verbal working memory 1 backtest: The control group committed no errors of commission and omission, whereas the mean number of errors of commission, omission, and total errors of commission and omission was highest in T0 group, which reduced on follow-ups. However, between-group differences were statistically non-significant.

Verbal working memory 2 backtest: The control group had committed fewer errors of omission, commission, and errors of omission and commission than the T0 group. The difference reached statistical significance in omission (P<.001, effect size. 22) and total errors of omission and commission (P<.001, effect size. 22), with a moderate effect size. These significant differences did not persist during the comparison of control, T1, and T2.

Visual working memory 1 backtest: The control group had committed fewer errors of omission, commission, and total errors of omission and commission than the patient groups (T0, T1, and T2). These differences were statistically significant.

Visual working memory 2 backtest: The control group had committed fewer errors of omission, commission, and the total errors of omission and commission than the T0 group. These differences were statistically significant (P <.001 for each parameter) with large effect sizes. On follow-up assessments, the mean number of errors of omission and total errors of omission and commission remained significantly higher at T1 and T2 when compared with control (P <.001 for each comparison, with large effect sizes) Table 3.

Table 3

Comparison of cognitive function of patients and healthy control at baseline, 3 months and 6 months follow-up

Parameter	Control (n=20) [mean (SD)]	T0 (n=60) [mean (SD)]	T1 (n=24) [mean (SD)]	T2 (n=17) [mean (SD)]	Baseline			3 months			6 months
					F	Partialη2 (adj)	P	F	Partialη2 (adj)	P	F	Partialη2 (adj)	P
Standard Progressive MatrixIQ	110.1 (10.3)	100.8(15.4)	103.2 (10.5)	104.7 (15.1)	3.0	.07	.03	2.3	.09	.09	1.6	.04	.22
Wisconsin Card Sorting Test
Total correct	47.8 (6.6)	42 (12.7)	47.7 (8.8)	46 (1.9)	1.3	.01	.28	.2	.06	.9	.1	.07	.95
Total error	16.2 (6.8)	23.2 (10.7)	16.3 (8.8)	16.5 (8.6)	3.1	.08	.03	.3	.05	.8	.11	.08	.95
Perseverative error	8.7 (4.9)	12 (6.3)	7.8 (5.1)	8.1 (5.2)	3.1	.08	.03	.2	.06	.89	.4	.05	.79
Non-perseverative error	7.5 (3.8)	9.1 (5.4)	8.5 (5)	8.5 (4.7)	.5	.04	.7	.4	.05	.79	.4	.05	.79
Conceptual level responses	43.1 (9.7)	34 (14.2)	42.6 (13.1)	42.3 (12.7)	3.1	.08	.03	.32	.05	.81	.17	.07	.92
Iowa Gambling Task
Net total score	46 (18.4)	2.5 (18.1)	47.2 (29.3)	44.6 (30.6)	85.8	.52	<.001*	1.5	.01	.23	.03	.03	.86
Total money earned	1214.3 (1286.3)	−625.2 (1362.7)	998.9 (1664.7)	857.4 (1748.9)	28.1	.26	<.001*	1.9	.02	.05	.53	.01	.47
Trail A (time in sec)	50.1 (14.7)	75.1 (25.7)	68 (28.9)	58.3 (19.8)	16.9	.17	<.001*	6.2	.11	.02	2.3	.03	.14
Trail B (time in sec)	70.7 (22.9)	116.8 (51.4)	96.6 (42.3)	83.2 (22.8)	14.9	.15	<.001*	5.9	.11	.02	3.8	.07	.06
Verbal Working Memory 1 back
Omission	0 (0)	.4 (.6)	.1 (.2)	.1 (.2)	8.2	.08	.005	7.1	.13	.01	1.1	.001	.31
Commission	0 (0)	.2 (.7)	0 (0)	0 (0)	1.9	.01	.17	7.1	.13	.01	-	-	-
Errors of omission and commissions	0 (0)	.6 (1.1)	.2 (.4)	.2 (.4)	6.5	.07	.01	9.5	.17	.004	3.5	.06	.07
Verbal Working Memory 2 back
Omission	.8 (.8)	2.1 (1.1)	.5 (.7)	.6 (.7)	23.7	.22	<.001*	7.6	.14	.009	.52	.01	.48
Commission	.3 (.5)	.6 (.9)	0 (0)	0 (0)	1.5	.006	.23	3	.05	.09	7.7	.15	.009
Errors of omission and commissions	1.1 (.8)	2.6 (1.4)	1.1 (2.2)	1.1 (2)	23.7	.22	<.001*	1.3	.01	.26	0	.03	1
Visual Working Memory 1 back
Omissions	.3 (.5)	1.9 (1.1)	1.4 (.71)	1.4 (.7)	41	.34	<.001*	45	.52	<.001*	32	.45	<.001*
Commissions	.1 (.2)	1.3 (1.1)	.5 (.5)	.4 (.5)	24.5	.23	<.001*	8.9	.16	.01	8.8	.17	.01
Errors of omission and commissions	.4 (.6)	3.2 (1.6)	2.1 (1.0)	2 (.9)	60.9	.43	<.001*	45.6	.52	<.001*	43.5	.53	<.001*
Visual Working Memory 2 back
Omission	.7 (.5)	2.9 (1)	2.2 (.6)	2.2 (.6)	89.8	.53	<.001*	97.1	.7	<.001*	72.4	.65	<.001*
Commission	.5 (.5)	1.9 (1.1)	.4 (.6)	.3 (.6)	34.4	.30	<.001*	6.7	.12	.01	.59	.01	.45
Errors of omission and commission	1.2 (.7)	4.9 (1.8)	2.5 (1)	2.5 (.9)	85.5	.52	<.001*	65.3	.61	<.001*	29.3	.43	<.001*

*Significant after Bonferroni correction. Effect size estimation: Partialη2; 0.01-0.09 small effect size; 0.09-0.25 medium effect size; >0.25 large effect size

Comparison of baseline (T0) neurocognitive functions of completers and non-completers: Baseline performance in cognitive function tests was compared between the patients who completed at least one follow-up (n = 24) and those who did not (n = 36). There was no statistically significant difference in any neurocognitive test performance between these two groups Supplementary Table 1.

Supplementary Table 1

Comparison of cognitive function between two subgroups of patients (patients who followed up at least once and who did not follow-up)

Parameters	F/up grp [mean(SD)] (n=24)	No F/up grp [mean(SD)] (n=36)	t	df	P
Standard Progressive MatrixIQ	97.3 (11.5)	103.7 (17.3)	−1.6	58	.12
Wisconsin Card Sorting Test
Total correct	39.2 (10.9)	44.1 (13.5)	−1.5	58	.14
Total error	24.8 (10.9)	21.8 (10.7)	1.1	58	.3
Perseverative error	13.5 (10.8)	10.8 (6.4)	1.6	58	.11
Non-perseverative error	11.3 (7.2)	11 (6.9)	.16	58	.88
Conceptual level responses	32.1 (14.4)	35.8 (14.2)	−.98	58	.33
Iowa Gambling Task
Net total score	2.4 (21.8)	2.6 (15.5)	.03	58	.98
Total money earned	882.7 (1333.5)	−453.5 (1373.3)	−1.2	58	.23
Trail A (time in sec)	85.8 (28.2)	67.9 (21.4)	2.8	56	58
Trail B (time in sec)	126.9 (56.3)	110 (47.5)	1.3	56	58
Verbal Working Memory 1 back
Omission	.5 (.6)	.4 (.6)	.81	58	.42
Commission	.2 (.7)	.2 (.7)	−.08	58	.94
Errors of omission and commissions	.7 (1.1)	.6 (1.1)	.43	58	.67
Verbal Working Memory 2 back
Omission	2.3 (1)	1.9 (1.2)	1.3	58	.21
Commission	.6 (.8)	.5 (.9)	.24	58	.81
Errors of omission and commissions	2.9 (1.5)	2.4 (1.3)	1.2	58	.24
Visual Working Memory 1 back
Omissions	2 (1.2)	1.8 (1)	.84	58	.40
Commissions	1.6 (1.1)	1.1 (1.1)	1.9	58	.07
Errors of omission and commissions	3.7 (1.6)	2.9 (1.5)	1.9	58	.06
Visual Working Memory 2 back
Omission	3.0 (1)	2.8 (1)	1.1	58	.26
Commission	2.3 (1.1)	1.7 (1)	2.1	58	.04
Errors of omission and commission	5.3 (1.7)	4.6 (1.8)	1.6	58	.11

F/Up – Follow-up; No P value is significant post Bonferroni correction

DISCUSSION

This study intended to examine: (a) the change of cognitive functions among patients with OD on BNX and (b) the difference between the neurocognitive functions between the patient group and the HC group across three different time points. The comparisons between the patients and HCs were adjusted for age and years of education. We adjusted the level of statistical significance for multiple comparisons. We tested for several neurocognitive domains: general intellectual functioning, verbal and visual working memory, attention and vigilance, executive function, decision making, impulsivity, risk-taking behavior, and psychomotor speed. We excluded comorbid illicit drug and alcohol dependence, HIV, head injury, and epilepsy to minimize the potential confounding effects of these conditions.

We observed a significant improvement in the performance of visual NBT after 3 months of initiation of buprenorphine (naloxone) treatment. Following 6 months of treatment, additional performance improvements were seen in several other tests such as WCST (perseverative errors), IGT, TMT-A, and verbal NBT. The test results suggested positive changes in cognitive flexibility, reduced impulsive decision making, improved attention, working memory, and psychomotor speed. A previous prospective study by Rapeli et al.[15] showed a significant improvement in the letter-number sequencing scores (assesses verbal working memory) between 6 and 12 months of buprenorphine treatment. The authors did not observe any significant differences in cognitive flexibility, attention and vigilance, and information processing. The set of cognitive tests administered in the study were different from ours. The mean dose of buprenorphine was much higher than in this study. Another study from the same group of authors did not show any improvement of working memory across two-time points of treatment. However, that study attributed the persistent cognitive impairment to benzodiazepine abuse.[24] A cross-sectional study, too, showed that the buprenorphine group performed better than the actively dependent group in TMT-A and B.[14]

We want to discuss the potential for practice effects in our results. In our opinion, the practice effects, albeit possible, might have minimally contributed to the improvement in the cognitive functions among the patient group. First, the practice effects are prominent in the early phases (before 3 months) and during high-frequency testing schedules (weekly).[25] Testing for our study was done at 3-monthly intervals. For TMT, practice effect becomes significant only after repetitive trials within a short period; TMT-A is more vulnerable to practice effect than TMT-B, which required more than three administrations in short intervals to produce a significant practice effect.[26] Second, most of the study participants needed a full explanation and introduction to tests during the repeat measurements. This observation indicated that the participants would be unlikely to have growing familiarity with the tests or recall effects. However, we agree that the effects of reduction of anxiety due to familiarity with the testing environment and regression to mean might still influence the test results. Third, practice effects have been seen to vary according to subjects’ age and education status.[2728] We have adjusted for the potential effects of these variables in our analysis. Fourth, although there are no head-to-head comparisons available, there is some evidence to suggest that tests for executive functions (e.g., WCST) and decision making (e.g., IGT) may be more susceptible to practice effects than the tests for working memory (e.g., verbal and visual N-back tests).[2529] Out of the five parameters measured in the WCST, only one (perseverative error) showed a significant change over time. Had practice effect been the most likely explanation for this improvement, we should have observed changes in other parameters as well. Fifth, the effect sizes of difference of means for all cognitive tests between the first and third assessments were large (>0.8). This large effects seemed unlikely to be solely explained by practice effects. Finally, the practice effect has been largely demonstrated among healthy participants. There is evidence to suggest those with pre-existing brain dysfunctions may not be able to learn from their mistakes.[253031] So it is unlikely that the patient group, with their considerable baseline cognitive deficit, would solely improve their performance due to practice effect. We think periodic tests in the control group could have thrown more lights on the potential practice effects and their impact on the improvement in neuropsychological functioning.

We also observed that the patients performed significantly worse at the baseline than the HCs in IGT, TMT-A, and B, and visual and verbal NBT. These test results indicated impaired decision making, attention, visual search, psychomotor speed, and working memory in the patient’s dependent on opioids. The profile was similar to those found in meta-analyses comparing OD and controls.[3233] The effect sizes were small to modest, which was also similar to the study by Wollman et al.[32] Interestingly, at 3 and 6 months of treatment, a significant difference between the BNX and HC was observed only for visual working memory. Previous studies reported deficits of working memory in patients on buprenorphine.[121424]

Limitations

Although we attempted to reduce the influence of confounders and effect modifiers in estimating the group effect (BNX vs HC), we might have missed several unmeasured confounders, such as the presence and severity of tobacco and cannabis dependence. About 62% of patients dropped out of the treatment or relapsed to illicit opioids and could not be included in the analysis. The drop-out rate was relatively higher than that was found in the recently published systematic review of observational studies.[34] However, the group retained in the treatment did not differ significantly from the drop-out group except for the occupational status and duration of OD. We could not draw any causal inference between BNX and the significant improvement of the cognitive functions because of a lack of a comparison group. An experimental study with a random allocation of participants to BNX and other treatments (e.g., antagonist or no treatment) could answer this question. Our study’s generalizability was limited because the sample was drawn from a single and specialized treatment center. Finally, the sample size for our study was calculated using attention and working memory as the outcome variable of interest. However, we examined changes in various other domains of cognitive functions (e.g., executive function and cognitive flexibility). The study might not be adequately powered to detect significant differences in paired observations of those domains.

IMPLICATIONS

Our study had the following clinical, research, and policy implications. Patients on medications for OD had significant improvement in multiple cognitive domains within the first 6 months of initiation, and a higher duration of BNX treatment seemed to be associated with greater improvements. This result supporting BNX should be continued for a longer duration, and premature termination could result in relapse and poor outcome. After short-term medication treatment (up to 12 weeks), relapse rates ranged from 50% to 80%, and poor cognitive functions might actually increase this risk.[3536] Besides, the findings dispelled the fear of buprenorphine (naloxone)-related cognitive dysfunctions because at the end of 6 months, except for visual working memory, the performance of the BNX group was not significantly different from HCs. The significant positive change in the neurocognitive functions among patients on BNX might also be reflected in the improvement of day-to-day functioning (e.g., driving). Previous studies, too, indicated similar conclusions.[3738] Finally, our results suggested a possibility of “brain recovery” in patients on a stable dose of BNX. In the future, researchers should try to address the question of the course of cognitive functions in a larger sample, powered to detect differences in all cognitive domains. The follow-up duration might be extended to see whether cognitive functions improve further with time. Research should also combine cognitive and functional outcomes to have more clinically relevant conclusions. To examine the possibility of “brain recovery,” structural and functional neuroimaging should also be combined with neurocognitive functions. At the policy level, our study indicated against time-limited medications for opioid use disorders. Higher cognitive dysfunctions at the initial stage would mean the treatment and service strategies should adapt to patients’ needs (e.g., provision of cognitive exercises and higher flexibility in service delivery).

In sum, this study found a significant improvement in cognitive flexibility, decision making, attention, working memory, and psychomotor speed among patients on a stable dose of BNX. Except for visual working memory, the BNX group did not differ from HC after 6 months.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.