Neuropsychological comparisons of cocaine versus methamphetamine users: A research synthesis and meta-analysis

ABSTRACT

Background: Previous meta-analytical research examining cocaine and methamphetamine separately suggests potentially different neuropsychological profiles associated with each drug. In addition, neuroimaging studies point to distinct structural changes that might underlie differences in neuropsychological functioning. Objectives: This meta-analysis compared the effect sizes identified in cocaine versus methamphetamine studies across 15 neuropsychological domains. Method: Investigators searched and coded the literature examining the neuropsychological deficits associated with a history of either cocaine or methamphetamine use. A total of 54 cocaine and 41 methamphetamine studies were selected, yielding sample sizes of 1,718 and 1,297, respectively. Moderator analyses were conducted to compare the two drugs across each cognitive domain. Results: Data revealed significant differences between the two drugs. Specifically, studies of cocaine showed significantly larger effect-size estimates (i.e., poorer performance) in verbal working memory when compared to methamphetamine. Further, when compared to cocaine, methamphetamine studies demonstrated significantly larger effect sizes in delayed contextual verbal memory and delayed visual memory. Conclusion: Overall, cocaine and methamphetamine users share similar neuropsychological profiles. However, cocaine appears to be more associated with working memory impairments, which are typically frontally mediated, while methamphetamine appears to be more associated with memory impairments that are linked with temporal and parietal lobe dysfunction.

KEYWORDS:

• Cocaine
• memory
• meta-analysis
• methamphetamine
• motor processes
• neurocognition
• neuropsychology
• stimulant dependence

Psychostimulants, such as cocaine and methamphetamine, are among the most commonly abused substances in the United States. A recent review reported prevalence rates for lifetime use of cocaine and methamphetamine as high as 14.7% and 8.5%, respectively (1). At low doses, both substances lead to increased vigilance, arousal, and attention (2). However, repeated and prolonged use of each substance has been associated with physiological consequences resulting in vascular and central nervous systems changes (2,3). While both drugs increase extracellular dopamine concentrations, they have been found to differ in their pharmacological mechanism (4) and therefore their potential impact on the brain. These distinctions are likely to translate into differences in neuropsychological functioning between the two groups of users (5,6).

Animal studies present strong evidence for the distinct pharmacologic effects of cocaine versus methamphetamine. Those studies demonstrated that, compared to cocaine, methamphetamine was associated with a significantly longer half-life and had a more substantial neurotoxic effect on striatal dopamine transporters (7,8). Additionally, Fleckenstein and colleagues (9) showed that multiple high doses of methamphetamine significantly decreased the efficiency of striatal dopamine transporters, while high-dose cocaine administrations had little to no impact. Similarly, using functional magnetic resonance imaging (fMRI), Taheri and colleagues (10) found that cocaine and methamphetamine induced opposite changes in the striatum of rats. The authors reported that methamphetamine resulted in a prolonged negative blood-oxygen-level-dependent signal, while cocaine yielded a positive signal.

Pharmacological distinctions between cocaine and methamphetamine have also been noted in humans. A positron emission tomography study found that compared to cocaine, methamphetamine showed a significantly prolonged rate of clearance in the striatum (4). Winhusen and colleagues (11) reported that compared to cocaine, methamphetamine had an accelerated and more widespread distribution across brain regions. The authors suggested that pharmacokinetic differences play a role in the observed neurotoxic impact on dopamine cells. These distinct patterns of brain changes are likely to be associated with different patterns in neuropsychological functioning.

There are many studies that have examined cognition in cocaine and methamphetamine users separately compared to their respective controls. Previous single-drug meta-analyses, such as the ones conducted by Scott and colleagues (12) on methamphetamine and Potvin and colleagues (13) on cocaine, have reported on the different cognitive domains impacted by each drug. Although these studies provided clarification as to the cognitive weaknesses associated with cocaine and methamphetamine use separately, they do not offer direct comparisons. Subsequently, these analyses do not provide insight on how the two drugs may differ in relation to cognitive deficits. Qualitatively reviewing the differences in effect-size calculations between these two meta-analyses suggests that each stimulant may be associated with a somewhat distinct cognitive profile.

There have been few primary studies that compare the neurocognitive differences associated with cocaine and methamphetamine (or amphetamine) dependence directly, and the results from those studies have been inconsistent. For example, Ersche and colleagues (5) found that cocaine users made significantly more perseverative errors than did amphetamine users on the Probabilistic Reversal Learning Task. These findings differed from Simon and colleagues (6) who found that methamphetamine users made significantly more perseverative errors than did cocaine users on the Wisconsin Card Sorting Test (WCST). In contrast, van der Plas and colleagues (14) found no differences between cocaine and methamphetamine users on WCST perseverative errors. In addition, several studies have suggested that both cocaine-dependent and methamphetamine-dependent individuals struggle significantly on measures of decision-making ability, as measured by the Iowa Gambling Task (IGT). Bechara and Martin (15) showed that methamphetamine users were more impaired than cocaine users on the IGT. However, van der Plas and colleagues (14) reported that the two drug groups did not differ significantly on IGT performance, despite both drug groups differing significantly from their respective control groups. Discrepant findings have also been demonstrated in working memory. Simon and colleagues (6) concluded that cocaine users showed unique difficulties on Digit Span-Backward that were not apparent in the methamphetamine sample. This finding differs with van der Plas and colleagues (14) who showed that methamphetamine users made more errors on the Tic-Tac-Toe working memory task than did cocaine users. Similarly, Bechara and Martin (15) found that methamphetamine users performed worse on a Delayed Non-Match to Sample task. However, Verdejo-Garcia and colleagues (16) found no differences between cocaine and methamphetamine users on a N-Back working memory task. Although these findings are inconsistent, they suggest the potential for distinct substance-related neurocognitive weaknesses in addition to overlapping deficits.

The potential for unique neuropsychological profiles is also supported by a recent structural neuroimaging meta-analysis on gray matter differences between cocaine and methamphetamine use conducted by Hall and colleagues (17). Results showed that while cocaine and methamphetamine users shared similar structural profiles, each substance was associated with unique regions of gray matter abnormalities. More specifically, cocaine users displayed greater gray matter reductions in the right insula and left frontal gyrus when compared to methamphetamine users. In contrast, methamphetamine users were shown to have significantly larger reductions in gray matter in the right inferior parietal lobe and the left temporal gyrus. Differences between the two drugs have also been observed in other neuroimaging studies. For example, research has shown significantly lower global cerebral blood flow in methamphetamine-dependent versus cocaine-dependent alcoholics (18). Cocaine and methamphetamine use has also been associated with regional differences in gray matter volumes (19–21). A recent qualitative article reviewing structural differences between cocaine and methamphetamine users suggested that cocaine users may show more reductions in gray matter in the temporal cortex when compared to methamphetamine users (22). Bartzokis and colleagues (23), who conducted a structural comparison between cocaine and methamphetamine reported temporal lobe changes only with cocaine users. These differences further suggest that each substance may be associated with unique neuropsychological distinctions, as structural brain changes have been shown to correlate with cognitive ability (24,25).

Taken together, the extant literature suggests that these two psychostimulants may be associated with distinctions in neurocognitive functioning, yet definitive conclusions regarding these differences in cognitive profiles remain elusive due to inconsistencies in the research. Disparities in the primary research may also be significantly influenced by differences in sample and methodological characteristics such as test selection, comorbid substance use, duration of use, and length of abstinence. Therefore, the objective of this study is to meta-analytically investigate the neurocognitive performance associated with cocaine versus methamphetamine use. It was postulated that a meta-analysis would demonstrate that cocaine- and methamphetamine-use is associated with neuropsychological profiles comprised of both shared and unique neurocognitive weaknesses compared to controls. Fully exploring the profiles associated with the two drugs may possibly aid clinicians in tailoring treatment according to the cognitive difficulties unique to users of each drug, thus potentially improving treatment outcomes.

Method

Search strategies and data acquisition

The methodology of this study was conducted following the Meta-Analysis Reporting Standards (MARS) guidelines (26). Searches covered the content of databases up to June 2016. Two researchers (MH, SW) conducted independent searches in nine databases: 1) Academic Search Complete E-Journals, 2) ArticlesFirst, 3) CINAHL PLUS, 4) PapersFirst, 5) ProceedingsFirst, 6) ProQuest Dissertation and Theses, 7) PsycINFO, 8) PUBMED, and 9) Web of Science. In an attempt to conduct a comprehensive search of the literature with a special emphasis on finding tests associated with standard neuropsychological testing batteries (27), clinically relevant neuropsychological search terms were determined by the two independent searchers in consultation with clinical experts and in accordance with Neuropsychological Assessment and the Compendium of Neuropsychology Tests (27,28). Search strategy and terms were independently determined by searchers to minimize potential study collection bias. Databases were searched using truncated terms such as “neuropsychology,” “neurocognitive,” “cognitive,” combined with terms related to neurocognitive domains such as “attention,” “motor,” “processing speed,” “memory,” “visuospatial,” or “executive functioning.” Databases were then searched using the truncated and related terms combined with substance related terms such as “crack,” “cocaine,” “methamphetamine,” or “amphetamine.” A specialist with professional experience in database searching (OA) created a third independent and extensive search based on the PsycINFO database algorithm to identify all abstracts included in PsycSCAN: Neuropsychology. Searches were presented to the PsycINFO staff for potential amendments, and the feedback was incorporated. Terms included controlled vocabulary (e.g., The Thesaurus of Psychological Index Terms) combined with free-text words from the title and abstract. A complete example of the search strategy and terms is published in Stephan and colleagues (29), however, cocaine and methamphetamine terms were used instead of alcohol terms. Results were combined to remove duplicates, and sorted into four categories: 1) Relevant (studies that were highly likely to be included), 2) Irrelevant (studies unrelated based upon title and abstract, such as animal or case studies), 3) Unknown (studies that required full-text examination before inclusion), and 4) Reviews (articles that may be utilized to identify additional sources of data; see Figure 1).

Figure 1. Flow chart of citations through searching, sorting, and consensus.

Inclusion and exclusion criteria

Studies were included based on the following a priori inclusion criteria: 1) the participants were identified in the primary studies as being current or lifetime psychostimulant abusing or dependent patients, 2) psychostimulant groups were compared to an appropriate control group (i.e., drug naive or limited use) and matched on at least age, education, and/or IQ, 3) data were provided on neuropsychological tests, 4) sufficient information was provided to calculate effect sizes, and 5) the article was published in English. Studies were excluded if 1) a significant proportion (>25%) of the control or psychostimulant group had comorbid conditions known to impact cognitive functioning (e.g., human immunodeficiency virus, hepatitis C, or traumatic brain injury), 2) the neuropsychological assessment was conducted using unstandardized procedures, 3) the majority of either group had current poly-substance dependence (other than nicotine or caffeine), and/or 4) the authors characterized their psychostimulant group as meeting criteria for major psychiatric disorder (e.g., psychotic disorder). Finally, it is important to note that studies were not excluded if the stimulant users had a history or current diagnosis of abuse of substances other than the stimulants.

Coding procedures

Fifty-four cocaine articles (5,6,14,30–80) and 41 methamphetamine articles (5,6,14,81–118) were selected for coding. Two independent reviewers coded all articles for the necessary neuropsychological data, as well as relevant demographic and drug-related variables. Prior to data being entered into the Comprehensive Meta-analysis Version 2 (CMA), a third independent reviewer assessed the coding for accuracy by randomly selecting and recoding five articles and examining potential outliers in the data.

Effect-size calculations and measures of heterogeneity

CMA was used to calculate effect-size estimates for Hedges’ g (119). Similar to other neuropsychological meta-analyses (13,120), the current study utilized Cohen’s effect-size ranges to describe magnitude of effect (121). The Q and I² statistics were used to assess for homogeneity of effect-size estimates across studies, while Τ² was utilized to estimate the variance of the true effect size across the population.

When multiple tests measuring the same construct were reported in a single study, composites were constructed in order to avoid violating the assumption of independence. The choice of domain in which to place a test consisted of a two-step process. Two independent coders examined the manuscripts for the author’s categorization of the tests in relation to the domain of functioning. If no such information was located or if different articles placed tests in different domains, then the coders independently consulted two references, the Compendium of Neuropsychology Tests and Neuropsychological Assessment (27,28) in order to classify the test into the appropriate domain.

All analyses were conducted using the random-effects model approach. In order to more accurately assess verbal memory, studies were separated into those that measured non-contextual memory (i.e., list-learning tests) and contextual verbal memory (i.e., story memory tests). Additionally, a subsample of the cocaine studies (k = 9) identified the participants as crack or crack/cocaine users. Similarly, a small proportion (k = 8) of the methamphetamine studies reported that the patients were amphetamine users. Each subsample was compared to their respective cocaine or methamphetamine samples. This revealed no differences in effect sizes and therefore, these subsamples were included in their respective group. Studies were classified as acute-users versus non-acute users based on the description of the participant’s last use by the primary authors. In order to assess potential differences between the acuteness (one week or less since last use or positive toxicology) and cognitive functioning, subgroup analyses were conducted comparing the two groups (acute versus non-acute). Additionally, length of use and length of abstinence were coded as potential moderating variables to analyze heterogeneity when the number of studies was greater than ten (122). Finally, publication bias was evaluated for each cognitive construct assessed using funnel plots and Duval and Tweedie’s Trim and Fill method (123).

Results

A total of 95 studies were included in this analysis (see Tables 1 and 2). Cocaine and methamphetamine users were compared on 15 different neuropsychological domains (see Table 3).

Table 1. Demographic and clinical characteristics of included cocaine studies.

Study	Cocaine Users (n)	Age	Education	Control (n)	Age	Education	Use	Abstinence	Domains Extracted
Abi-Saab (30)	93	34.8	12.2	16	32.1	14.6	12.3	Acute	CMS, EF, IL, MMS, PMS, VF, VRM, VWM
Adinoff (31)	13	41.2	–	15	35.0	–	–	21	EF
Albein-Urios (32)	24	33.2	9.6	20	28.6	10.6	4.42	14	A, EF, IQ, VWM
Albein-Urios (33)	36	32.0	10.3	34	30.5	10.5	4.36	70	APS, EF, IQ, VWM
Balconi (34)	34	54.8	13.2	39	56.4	13.9	12.3	15	EF
Beatty (35)	23	34.0	13.7	22	35.5	14.0	7.10	21	EF, IL, IQ, PMS, VM, VRM, VWM
Berry (36)	16	34.5	14.0	21	33.2	14.1	–	14	A, EF IL, IQ, PMS, VF, VSP,
Bolla (37)	30	33.5	12.0	21	33.9	12.6	6.70	30	CMS, EF, IL, IQ, MMS, PMS, VF, VRM, VM, VSP
Browndyke (38)	37	35.5	12.3	13	30.5	14.0	12.4	14	A, CMS, MMS
Cheney (39)	16	37.4	12.3	16	38.0	12.9	8.06	35	A, IQ, VM, VRM, VSP
Cunha (40)	15	25.6	10.5	15	26.6	11.2	7.80	15	EF, IQ, L
Cunha (41)	58	31.2	11.7	32	26.7	10.7	10.7	12	A, APS, EF, IQ, L
De Oliveira (42)	20	25.9	10.9	18	24.2	11.5	2.90	180	A, EF, IL, L, PMS, VF, VRM, VWM
Di Sclafani (43)	19	39.2	12.8	19	35.3	15.3	12.3	72	A, EF, VSP
Di Sclafani (44)	20	40.0	13.0	29	37.0	15.0	14.0	42	A, CMS, PMS, VF, VSP
Elman (45)	27	37.8	12.8	33	31.1	15.7	15.1	Acute	VSP
Ersche (5)	27	32.3	–	12	35.1	–	8.70	Acute	EF, IQ
Fernandez-Serrano (46)	42	28.9	11.8	65	30.1	17.0	4.13	239	EF
Gillen (47)	19	34.3	13.6	16	30.9	14.4	6.30	Acute	A, EF, IL, L, PMS, VF, VM, VRM, VSP, VWM,
Goldstein (48)	30	42.6	12.6	42	40.9	13.6	18.7	Acute	IQ, L
Hanlon (49)	24	37.0	–	25	36.2	–	10.2	243	EF, IQ
Hoff (50)	38	34.3	13.2	54	32.1	15.4	7.30	24	EF, IL, IQ, L, MMS, PMS, VF, VM, VRM
Kalapatapu (51)	20	56.8	13.5	20	55.4	14.7	–	–	A, EF, IQ, PMS, VWM
Kelley (52)	12	36.5	13.1	12	33.1	16.4	–	Acute	APS, EF, IQ, L, VF, VM, VRM, VSP
Kjome (53)	66	36.7	–	20	33.9	–	–	–	EF
Lane (54)	18	39.2	12.6	15	34.0	16.0	14.6	–	EF
Lane (55)	15	38.4	13.1	18	35.2	15.4	16.1	–	EF
Liu (57)	37	40.5	12.9	32	32.9	13.3	13.6	Acute	A, EF
Madoz-Gurpide (58)	24	35.9	–	27	33.4	–	–	Acute	A, EF, VWM
Meade (59)	53	45.8	12.0	32	44.7	13.8	17.9	–	A, CMS, EF, IL, IQ, L, PMS, VF, VM, VRM, VWM
Mittenberg (60)	16	29.3	14.1	16	29.4	14.6	5.00	20	IQ, L, VRM, VSP
Moeller (61)	30	36.5	–	25	29.6	–	–	Acute	EF
Moeller (62)	18	33.1	–	18	30.3	–	9.70	Acute	A, EF
Moeller (63)	16	48.0	13.0	23	42.7	13.9	18.3	Acute	IL, IQ, VRM
O’Malley (64)	20	27.7	13.3	40	27.6	13.2	4.15	23	EF, IL, IQ, L, PMS, VRM, VM, VSP
Patzelt (65)	45	41.0	12.9	41	40.0	14.5	14.7	Acute	EF
Ranlett (66)	20	36.3	13.2	18	36.3	13.8	13.0	34	A, IQ, VRM
Ray Li, 2006	18	39.1	11.8	41	34.1	12.9	10.9	14	EF
Robinson (67)	30	28.7	13.0	30	29.0	13.0	4.10	146	A, CMS, IL, IQ, MMS, PMS, VSP, VWM
Rosselli (68)	61	26.1	10.4	63	25.6	10.5	10.2	78	EF, IL, IQ, L, PMS, VF, VM, VRM, VSP, VWM
Rosselli (69)	42	34.1	11.4	17	34.5	12.2	16.0	60	A, EF, L, PMS, VF, VM, VRM, VSP, VWM
Selby (70)	60	33.1	11.2	138	33.9	11.4	12.5	1075	CMS, EF, IQ, PMS, VF, VM, VRM, VSP
Simon (6)	40	38.0	12.9	40	35.0	13.7	–	Acute	APS, EF, IL, IQ, PMS, VF, VM, VRM, VWM
Smelson (71)	35	36.1	–	17	40.0	–	–	Acute	CMS, EF, IL, PMS, VWM
Stout (72)	12	36.9	–	14	30.0	–	–	–	EF, IQ
Tucker (73)	60	35.3	12.3	13	33.5	13.0	11.9	Acute	A, CMS, EF, IL, MMS, PMS, VRM, VWM
Vadhan (75)	22	39.3	12.4	21	38.9	13.3	21.0	Acute	APS, EF
Vadhan (74)	22	40.3	13.1	24	36.9	15.4	20.5	Acute	EF, IQ
van der Plas (14)	38	34.5	11.4	36	28.9	16.2	10.1	14	EF, VWM
van Gorp (76)	37	39.2	13.2	27	38.2	14.1	0.71	Acute	VM
Verdejo-Garcia (77)	12	36.8	13.3	14	30.9	13.6	8.50	Acute	EF, IQ
Vonmoos (78)	30	32.5	10.0	68	30.3	10.7	9.40	21	A, EF, IQ, VRM, WVM
Woicik (79)	21	40.9	12.5	64	38.7	13.6	18.5	15	CMS, EF, IL, IQ, L, MMS, PMS, VF, VRM, VWM
Woicik (80)	107	43.3	12.9	107	38.1	13.3	–	Acute	EF, IQ
Total	1,718	36.5	12.2	1,668	34.3	13.4	11.0	132.9

Note. n = sample size; Use = total duration of use in years; Abstinence = abstinence in days, Acute = toxicology was positive on the day of testing, or last use was one week prior to testing. All total averages are weighted by sample size. A = Attention; APS = Auditory Processing Speed; CMS = Coordinated Motor Speed; EF = Executive Functioning; IL = Incidental Learning; IQ = Intelligence Quotient; L = Language; MMS = Manual Motor Speed; PMS = Psychomotor Speed; VF = Verbal Fluency; VM = Visual Memory; VRM = Verbal Memory; VSP = Visuospatial Processing; VWM = Verbal Working Memory.

Table 2. Demographic and clinical characteristics of the included methamphetamine studies.

Study	Meth Users (n)	Age	Education	Control (n)	Age	Education	Use	Abstinence	Domains extracted
Ahn (81)	38	22.7	12.5	48	24.7	13.8	3.20	90	EF, IQ
Al-Zahrani (82)	56	33.3	–	100	31.2	–	–	–	APS, EF, IL, PMS, VF
Boileau (83)	16	27.8	12.1	14	29.7	15.5	5.10	19	CMS, EF, IQ, PMS, VWM
Bousman (84)	50	36.0	13.0	49	37.0	13.0	10.0	125	EF, IQ
Carter (85)	23	33.7	–	21	36.2	–	–	–	EF
Chang (86)	20	33.1	12.8	13	30.9	14.4	9.83	120	APS, CMS, EF, IL, PMS, VRM
Cherner (87)	54	36.3	12.6	46	35.1	12.2	12.0	129	A, CMS, EF, IL, IQ, PMS, VF, VM, VRM, VWM
Chung (88)	23	36.0	–	20	33.3	–	6.24	729	EF
Dean (89)	24	35.9	12.9	17	31.1	14.5	10.1	11	APS, CMS, EF, IL, MMS, VWM
Ersche (90)	25	37.0	–	27	35.1	–	16.9	Acute	EF, VM, VSP
Ersche (5)	30	37.4	–	13	35.1	–	15.8	Acute	EF, IQ
Gonzalez (92)	26	39.3	12.1	41	34.6	12.9	12.1	135	A, CMS, EF, IQ, L, VF, VM, VRM, VWM
Gonzalez (91)	16	35.4	11.7	19	31.1	14.9	16.7	29.9	EF, IQ
Han (93)	37	35.1	10.2	40	34.4	13.0	11.3	28	EF
Hekmat (94)	52	28.3	–	130	30.1	–	–	16	APS, EF, IL, PMS
Henry (96)	12	27.5	11.7	12	28.2	12.7	3.90	177	EF, IQ, VF, VRM
Henry (95)	16	36.7	13.6	18	36.1	13.9	10.3	290	EF, IQ
Hoffman (97)	41	37.8	12.3	41	35.1	13.5	2.55	196	APS, CMS, EF, IQ, PMS, VM, VRM, VSP
Hosak (98)	43	25.3	11.4	52	38.7	13.3	6.20	30	EF
Johanson (99)	16	32.0	–	18	31.0	–	10.3	90	A, CMS, EF, IL, L, MMS, PMS, VF, VRM
Kalechstein (101)	27	35.4	12.6	18	32.7	13.2	–	Acute	APS, EF, IL, IQ, PMS, VF, VM, VRM, VSP, VWM
Kalechstein (100)	25	34.8	12.7	19	32.5	13.7	–	Acute	EF, IQ
Kim (102)	29	36.6	10.7	20	33.2	15.1	5.35	600	EF
Kim (104)	24	37.7	10.5	21	39.1	12.4	13.6	20	EF
Kim (103)	28	40.2	14.3	27	38.9	14.8	13.9	19	EF
Marquine (105)	25	38	12.0	47	38.4	14.0	4.38	31	IQ
McCann (106)	22	33.4	13.0	17	34.7	14.5	8.44	867	APS, IQ, L, MMS, VRM
Mcketin (107)	15	22.0	–	9	22.0	–	–	–	A, IQ, L, VRM
Moon (108)	17	–	–	18	–	–	12.8	653	VM, VRM, VSP
Rapeli (109)	12	26.8	10.9	12	28.2	13.8	9.30	1206	A, APS, EF, IQ, VRM, VWM
Rendell (110)	20	27.5	11.6	20	28.2	12.6	3.85	177	A, EF, IQ, VF, VRM, VWM
Rogers (111)	18	35.5	–	26	37.4	–	13.4	Acute	EF, IQ
Segala (112)	78	23.6	12.9	102	25.2	13.3	3.60	817	A, EF, IQ
Sim (113)	41	31.9	13.2	39	29.6	13.3	–	Acute	EF, IQ, PMS
Simon (115)	65	32.0	12.4	65	31.0	12.8	11.2	Acute	APS, EF, IL, IQ, PMS, VF, VM, VRM, VWM
Simon (6)	40	32.0	12.5	40	30.0	13.0	–	Acute	APS, EF, IL, IQ, PMS, VF, VM, VRM, VWM
Simon (114)	25	33.9	12.7	24	33.3	15.1	7.63	Acute	APS, EF, IQ, PMS, VF, VM, VRM, VWM
van der Plas (14)	38	30.7	11.7	36	28.9	16.2	7.80	15	EF
Van Wyk (116)	17	–	–	18	–	–	–	30	A, VM
Wang (117)	26	27.3	9.23	39	30.0	9.77	2.31	14	EF
Woods (118)	87	37.8	12.5	71	36.1	12.7	12.6	120	EF, IQ, VRM
Total	1,297	32.9	12.2	1,427	32.3	13.4	9.0	223.5

Note. n = sample size; Use = total duration of use in years; Abstinence = abstinence in days, Acute = toxicology was positive on the day of testing, or last use was one week prior to testing. All total averages are weighted by sample size. A = Attention; APS = Auditory Processing Speed; CMS = Coordinated Motor Speed; EF = Executive Functioning; IL = Incidental Learning; IQ = Intelligence Quotient; L = Language, MMS = Manual Motor Speed; PMS = Psychomotor Speed; VF = Verbal Fluency; VM = Visual Memory; VRM = Verbal Memory; VSP = Visuospatial Processing; VWM = Verbal Working Memory.

Table 3. Breakdown of the measures included for each construct assessed.

Construct	Variable	Measure
Perseveration	Perseverative Errors or Responses	WCST
Inhibition	Time/Number Correct	Stroop; C-WIT
Mental Flexibility	Time	TMT-B
Decision Making	Net Score/Advantageous/Disadvantageous Choices	IGT; Decision Making Task
Verbal Working Memory	Number Correct	DS-B/T; Arithmetic; N-LS
Coordinated Motor Speed	Time	Grooved Pegboard
Manual Motor Speed	Time	Finger Tapping Test
Psychomotor Speed	Time	TMT-A
Incidental Learning	Number Correct	Digit Symbol
Verbal Fluency	Total Number	COWAT
Non-Contextual Verbal Memory	Delayed Memory	CVLT; HVLT; RAVLT; KAVLT
Contextual Verbal Memory	Delayed Memory	LM; BST
Auditory Processing Speed	Time/Number Correct	Color Naming, Word Reading
Attention	Total Correct	Digit Span Forward; PASAT
Visual Memory	Delayed Memory	BVMT; Rey-O; VR; DRMT; PRM; K-CFT

Note. WCST = Wisconsin Card Sorting Test; C-WIT = Color Word Interference Test; TMT-B = Trail Making Test part B; IGT = Iowa Gambling Test; DS-B/T = Digit Span Backward/Total; N-LS = Number-Letter Sequencing; TMT-A = Trail Making Test part A; COWAT = Controlled Oral Word Association Test; CVLT = California Verbal Learning Test; HVLT = Hopkins Verbal Learning Test; RAVLT = Rey Auditory Verbal Learning Test; KAVLT = Korean Auditory Verbal Learning Test; LM = Logical Memory; BSMT = Babcock Story Memory Test; PASAT = Paced Auditory Serial Attention Test; Delayed Pattern Recognition Memory; Rey-O = Rey–Osterrieth Complex Figure Test; VR = Visual Reproduction; BVMT = Brief Visuospatial Memory Test; RMT = Repeated Memory Test; Korean-Complex Figure Test.

Individual drug effect sizes

Most effect sizes within each of the drug classes were statistically significant (see Table 4). The only effect size that did not reach statistical significance was manual motor speed (g = .147, p = .111) for cocaine.

Table 4. Effect size estimates and subgroup analyses comparing cocaine and methamphetamine users.

		Effect size estimates					Test of heterogeneity			Subgroup analyses
Construct	k	g (SE)	95% CI	Z	P	I^2	Τ^2	Q	p	Q	df	p
Perseveration										1.853	1	0.173
Cocaine	12	.396(.090)	.220 to .571	4.413	.000	37.024	.034	17.467	.095
Methamphetamine	14	.557(.078)	.404 to .711	7.130	.000	32.223	.027	19.181	.118
Inhibition										3.187	1	.074
Cocaine	13	.330(.091)	.150 to .509	3.606	.000	44.108	.046	21.470	.044
Methamphetamine	9	.551(.084)	.387 to .716	6.567	.000	.000	.000	6.557	.585
Mental Flexibility										1.661	1	.197
Cocaine	18	.498(.083)	.335 to .662	5.984	.000	45.428	.054	31.152	.019
Methamphetamine	13	.360(.068)	.227 to .493	5.299	.000	0.000	.000	6.797	.871
Decision Making										.679	1	.411
Cocaine	9	.743(.118)	.512 to .973	6.310	.000	19.593	.024	9.949	.269
Methamphetamine	7	.562(186)	.197 to .926	3.016	.003	71.337	.163	20.933	.002
Verbal Working Memory										7.736	1	.005*
Cocaine	16	.650(.117)	.421 to .879	5.567	.000	68.642	.146	47.835	.000
Methamphetamine	10	.226(.098)	.033 to .418	2.297	.022	28.729	.027	12.628	.180
Coordinated Motor Speed										.229	1	.632
Cocaine	10	.474(.139)	.201 to .747	3.399	.001	69.069	.128	29.097	.001
Methamphetamine	6	.392(.100)	.195 to .588	3.898	.000	.000	.000	2.233	.818
Manual Motor Speed										2.160	1	.142
Cocaine	7	.147(.093)	−.034 to .329	1.592	.111	.000	.000	4.614	.594
Methamphetamine	3	.436(.173)	.097 to .775	2.519	.012	.000	.000	.074	.964
Psychomotor Speed										.011	1	.915
Cocaine	18	.348(.081)	.189 to .507	4.293	.000	43.734	.050	30.214	.025
Methamphetamine	10	.360(.078)	.208 to .513	4.630	.000	.000	.000	5.187	.818
Incidental Learning										.649	1	.420
Cocaine	12	.578(.097)	.387 to .768	5.938	.000	34.704	.038	16.846	.113
Methamphetamine	4	.696(.110)	.480 to .913	6.305	.000	.000	.000	2.691	.442
Verbal Fluency										.428	1	.513
Cocaine	14	.184(.091)	.006 to .363	2.023	.043	43.651	.048	23.071	.041
Methamphetamine	10	.289(.131)	.032 to .546	2.201	.028	63.834	.104	24.885	.003
Non-Contextual Del. Ver. Memory										.123	1	.726
Cocaine	13	.485(.125)	.240 to .730	3.875	.000	60.077	.119	30.058	.003
Methamphetamine	14	.535(.067)	.403 to .667	7.927	.000	.000	.000	7.418	.879
Contextual Del. Ver. Memory										4.565	1	.033*
Cocaine	7	.305(.129)	.051 to .558	2.356	.018	52.008	.058	12.502	.052
Methamphetamine	4	.726(.149)	.434 to 1.017	4.884	.000	.000	.000	1.413	.703
Auditory Processing Speed										.070	1	.791
Cocaine	5	.258(.130)	.002 to .514	1.976	.048	30.340	.026	5.742	.219
Methamphetamine	11	.219(.067)	.088 to .350	3.274	.001	.000	.000	8.867	.545
Attention										1.199	1	.273
Cocaine	10	.615(.172)	.278 to .951	3.582	.000	70.470	.204	30.478	.000
Methamphetamine	4	.347(.175)	.004 to .689	1.984	.047	36.850	.045	4.751	.191
Delayed Visual Memory										4.622	1	.032*
Cocaine	13	.302(.072)	.162 to .442	4.220	.000	12.786	.008	13.759	.316
Methamphetamine	9	.550(.091)	.373 to .728	6.078	.000	18.766	.014	9.848	.276

Note: k = number of comparisons in the analysis; g = Hedges’ g effect size; SE = Standard error for Hedges’ g; CI = confidence interval; z = z-score; p = level of significance; df = degrees of freedom; I² = amount of heterogeneity as a percentage of total variance; Τ² = estimate the variance of the true effect size; Q = Q-statistic utilized to measure heterogeneity.

Neuropsychological profile overlap

The following subgroup comparisons assessing various aspects of executive functioning were not statistically significantly different between the two drugs: mental flexibility, as measured by the Trail Making Test-B (TMT-B); inhibition, as measured by Stroop; decision-making, as measured by the IGT and the Decision Making Test; and perseveration, measured by the WCST. In addition, the Controlled Oral Word Association Test, which comprised the verbal fluency domain, was not significantly different between the two types of drug studies. The non-contextual verbal memory domain (i.e., list learning tasks) was not significantly different between the groups. Processing speed measures not statistically different were: Stroop Color Naming and Word Reading (CNWR), Trail Making Test-A (TMT-A), and the Digit Symbol Test (DST). Measures of attention were also found to not differ significantly between the two drug using groups. Finally, coordinated motor speed and manual motor speed were also not statistically different between the two groups of drug studies (All ps > .05).

Differences between methamphetamine and cocaine domains

Verbal working memory

Subgroup analysis revealed a significant difference between cocaine and methamphetamine users (Q = 7.736, df = 1, p = .005) on verbal working memory. More specifically, results indicated a significant and medium effect size for cocaine g = 0.650 (95% CI: [.421 to .879], z = 5.567, df = 15, p = .000), with high and significant heterogeneity I² = 68.642% (Q = 47.835, p = .000, Τ² = .146) and a significant and small effect for methamphetamine g = 0.226 (95% CI: [.033 to .418], z = 2.297, df = 9, p = .022), without significant heterogeneity I² = 28.729% (Q = 12.628, p = .180, Τ² = .027).

Contextual delayed verbal memory

Subgroup analysis for contextual delayed verbal memory revealed a significant difference between the groups (Q = 4.565, df = 1, p = .033). A significant and small effect for cocaine was observed g = .305 (95% CI: [.051 to .558], z = 2.356, df = 6, p = .018), without significant heterogeneity I² = 52.008% (Q = 12.502, p = .052, Τ² = .058) and a significant and large effect for methamphetamine was found g = 0.726 (95% CI: [.434 to 1.017], z = 4.884, df = 3, p = 0.000), without significant heterogeneity I² = 0.00% (Q = 1.413, p = .703, Τ² = 0.000).

Delayed visual memory

Subgroup analysis revealed a significant difference between cocaine and methamphetamine users (Q = 4.622, df = 1, p = .032) on measures of delayed visual memory. Results indicated a significant small effect for cocaine g = .302 (95% CI: [.162 to .442], z = 4.220, df = 12, p = 0.000), without significant heterogeneity I² = 12.786% (Q = 13.759, p = .316, Τ² = .008) and a significant and medium effect for methamphetamine g = .550 (95% CI: [.373 to .728], z = 6.078, df = 8, p = 0.000), without significant heterogeneity I² = 18.766% (Q = 9.848, p = .276, Τ² = .014).

Moderator analyses

Length of use and length of abstinence were not found to significantly predict any effect sizes in meta-regression analyses. Subgroup analyses results indicated that there were no differences between acute and non-acute users except on verbal fluency (Q = 15.9, df = 1, p < .000). Specifically, acute cocaine users were found to have a significantly larger effect-size estimate when compared to non-acute cocaine users.

Risk of publication bias

The Trim and Fill method (123) revealed possible overestimation of effect size in four domains: non-contextual delayed verbal memory (biased estimate g = .506 and unbiased estimate g = .427), mental flexibility (biased estimate g = .405 and unbiased estimate g = .284), decision-making (biased estimate g = .526 and unbiased estimate g = .447), and manual motor speed (biased estimate g = .211 and unbiased estimate g = .135). Funnel plots for each cognitive construct are available in supplemental material.

Discussion

The current study is the first to meta-analytically compare the neuropsychological functioning of cocaine versus methamphetamine users. Findings from this study suggest that cocaine and methamphetamine users share a similar neuropsychological profile. This is expected since both drugs impact monoamines, affect reuptake transporters, and release acetylcholine in the brain’s reward pathway. Additionally, both substances result in the downregulation of dopaminergic receptors (1). Users of both substances included in this meta-analysis differed significantly from their respective controls on all but one domain (manual motor speed for cocaine). Both cocaine and methamphetamine studies were found to have large effect-size estimates (i.e., poorer performance compared to controls) in the domain of decision-making, comprised primarily of the IGT. The IGT has been found to have high ecological validity and has also been shown to be predictive of real-life decision-making among substance-dependent individuals (124). This is consistent with research that has demonstrated that stimulant users struggle with areas of daily functioning that require intact decision-making skills related to obtaining employment, medical care, and continued sobriety (125). Users of both substances displayed significant and moderate effect-size estimates on the domains of executive functioning, which is generally consistent with previous single-drug meta-analyses (12,13). These findings may have important implications for clinicians. Patients may benefit from cognitive rehabilitation strategies focused on recovery and/or acquisition of skills related to executive functioning. Such rehabilitation has been shown to help patients with executive dysfunction following traumatic brain injuries (126). Specifically, in a sample of brain injured patients, Levine and colleagues (127) found that seven two-hour sessions of Goal Management Training (GMT) led to improvements on various neuropsychological tests measuring working memory, attentional ability, behavioral consistency, and problem-solving. Future research should investigate the possibility of improving treatment outcomes by utilizing cognitive remediation techniques such as those used in patients with traumatic brain injury.

Difficulties in processing speed have also been previously demonstrated meta-analytically in both methamphetamine and cocaine users (12,13). Specifically, Scott and colleagues (12) reported a medium effect size (e.g., d = .52) for the domain of processing speed, which included TMT-A, Stroop CNWR, and the Digit Symbol Test. Similarly, Potvin and colleagues (13) reported a comparable effect size in the speed of processing domain (e.g., d = .45), which included TMT-A, Stroop CNWR, DST, and Grooved Pegboard. In the current meta-analysis, each test was examined separately. By reducing heterogeneity, a better picture of the relationships between these different components of processing speed emerged. The results suggest that performance on DST may have had the largest impact on the summary effect size for processing speed in previous meta-analyses. In the current study, both cocaine and methamphetamine studies were shown to have similar and substantial effect sizes on the DST (g = .57 and g = .69, respectively), while the effect sizes for TMT-A and Stroop CNWR speed were minimal. It appears that the DST might be more sensitive to the impact of cocaine and methamphetamine than tests that are typically used to assess other aspects of processing speed. This sensitivity is likely due to the complex nature of this test. Therefore, complex tasks that require sustained and/or divided attention are likely to be challenging for patients with a history of stimulant abuse. One potential practical implication of this finding is that clinicians may consider providing information in smaller chunks in order to allow ample time for the processing and integration by patients.

The current analysis also found that cocaine and methamphetamine studies did not significantly differ on tasks designed to measure attention. Cocaine users were found to have a medium-to-large effect size (g = .61), a result consistent with Potvin and colleagues (13), who reported an effect size of d = .59 for their attention domain. Similarly, the effect size of methamphetamine users was relatively smaller (g = .35), a finding consistent with Scott and colleagues (12), who reported an effect size of d = .39. Research has indicated that attentional ability significantly impacts the success of treatment interventions (128). It has been suggested that adequate attentional capacity is essential to encode and process information and therefore, when information is inadequately processed, patients’ abilities to transform therapeutic information into behavioral change may be negatively impacted (128).

In addition to substantial overlap, this study also suggests distinctions between cocaine and methamphetamine users, which appear to correspond to their respective brain structural differences. Cocaine users exhibited a significantly larger effect-size estimate on verbal working memory when compared to methamphetamine users. Hall and colleagues (17) found that compared to methamphetamine users, cocaine users showed significantly more reduced gray matter in the left inferior and superior frontal gyri, which are important regions in the network system for working memory. Research has shown that scores on Digit Span backward significantly correlated with gray matter volume in the left inferior frontal gyrus (129). This region has been shown to exhibit increased cerebral blood flow during the rehearsal process, which is a component of working memory procedures (130). Research has also shown an association between the left superior frontal gyrus and working memory performance (131). Du Boisgueheneuc and colleagues (131) concluded that the left superior frontal gyrus appears more involved in higher order aspects of working memory performance, such as monitoring and manipulation. Similarly, Chen and colleagues (132) reported reduced activation of the left inferior frontal gyrus under a low-load working memory condition and reduced activation of the left superior frontal gyrus under the high-load working memory condition in patients with diabetes. These findings could help explain the verbal working memory weaknesses observed in cocaine-dependent individuals, which may have important clinical implications. Performance on verbal working memory tasks have been shown to correlate with treatment goals and relapse in substance users (128,133). Given that working memory processes have been postulated to be related to the regulation of unwanted craving (134), providers should be mindful that cocaine users may be especially susceptible to craving-induced relapse. Cognitive remediation tools such as Work Therapy and Neurocognitive Enhancement Therapy have been shown to improve working memory performance in psychiatric populations (135). Additional rehabilitation strategies such as GMT have been shown to be effective at improving working memory and other cognitive functions in substance using populations. Specifically, Alfonso and colleagues (136) demonstrated that compared to treatment as usual, seven weeks of GMT led to significant improvements in working memory, response inhibition, and decision-making in a population of alcohol and polysubstance abusers. Implementing these and other interventions may aid cocaine-using individuals with difficulties in working memory.

When compared to cocaine, methamphetamine users performed significantly worse on measures of delayed contextual verbal memory. Hall and colleagues (17) reported that when compared to cocaine users, methamphetamine users showed significantly more reductions in gray matter in the left superior temporal gyrus (L-STG), particularly Heschl’s gyrus, a region that might be considered part of a network system that is correlated with contextual verbal memory. Functional and structural studies have reported negative associations between activation and volume in the L-STG and performance on Logical Memory (137,138). Additionally, reduced cerebral perfusion in areas adjacent to the L-STG has been positively correlated with performance on the Babcock Story Memory Test (139). While these studies provide evidence for the involvement of the L-STG in contextual verbal memory, they do not specify the regions of the superior temporal gyrus that may be more or less involved during contextual verbal memory performance. However, by manually segmenting the L-STG, researchers have shown the left Heschl gyrus to correlate with Logical Memory performance (140). Interestingly, the researchers reported that performance on the California Verbal Learning Test (CVLT) did not correlate with this region. The latter finding may provide some clarification as to why the current analysis did not find significant differences between the drugs on non-contextual verbal memory tests. These findings suggest that while users of both drugs perform poorly on non-contextual verbal measures, methamphetamine users appear to perform significantly worse on measures of contextual verbal memory compared to cocaine users. This finding is important for treatment providers, since it appears that methamphetamine users may be less likely to benefit from the context of information presented in a verbal format. Additionally, research has shown contextual verbal memory performance to correlate with the amount of treatment-related information learned by patients at discharge (141). Therefore, providers may help patients, especially methamphetamine users, by implementing strategy training techniques designed to compensate for verbal memory difficulties (142). Specifically, Kaschel and colleagues (142) conducted a small study of brain-injured patients and found that 30 sessions of imagery-based mnemonics improved delayed recall of everyday relevant verbal material such as stories and appointments. Additionally, a recent study by Rupp and colleagues (143) compared the effectiveness of cognitive behavioral therapy (CBT) and CBT supplemented with cognitive rehabilitation in an alcohol dependent population. The authors found that CBT supplemented with 12 sessions of computer-assisted cognitive rehabilitation led to significant improvements in delayed verbal memory as well as other cognitive abilities. Finally, it is important for providers to be aware of the necessity of considering alternatives to verbally mediated instructions for both drug groups, but potentially more so for methamphetamine-dependent patients.

Methamphetamine users also performed significantly worse on measures of delayed visual memory when compared to cocaine users. This finding appears consistent with Hall and colleagues (17) who demonstrated that methamphetamine users have significantly greater reductions in gray matter in the right inferior parietal gyrus (particularly the supramarginal gyrus), which is a region potentially involved in the visual memory system. Specially, these regions have been shown to be involved in the processing of visually presented stimuli and the retrieval of spatial location (144,145). Melrose and colleagues (144) found that the right supramarginal gyrus was correlated with cerebral metabolism during performance of the Rey–Osterrieth Complex Figure Test (Rey-O). Further, a voxel-based lesion mapping study has shown these regions to be associated with Rey-O performance (146). Evidence from these studies suggests that the right inferior parietal lobe, particularly the supramarginal gyrus, appears to play a role in the processing of visually presented information. Importantly, visual memory performance has been shown to correlate with treatment engagement, as well as substance relapse (147). Cicerone and colleagues (148) suggested that memory strategy training, such as internalized strategies (e.g., visual imagery) and external memory compensations (e.g., notebooks) may improve and compensate for memory difficulties in individuals with acquired brain injury. It is possible that methamphetamine users may also benefit from such interventions.

Previous research has demonstrated that length of substance use and abstinence are significant predictors of level of cognitive functioning (149,150). However, results from the present analysis indicated that neither length of use nor length of abstinence were significant predictors of effect sizes. It is possible that the limited number of studies included in each analysis resulted in insufficient statistical power to detect moderating effects.

Despite careful control over the studies that were included, the current meta-analysis has several limitations. Specifically, given that some of the primary studies included participants with current or past substance abuse, it is difficult to rule out the confounding effect of additional substances. Additionally, this meta-analysis included only articles that were published in English. Considering the high rate of stimulant abuse in non-English-speaking countries (151), this inclusion criterion may limit generalizability. Females were significantly underrepresented which may limit generalizability across genders. Finally, considering the observed heterogeneity in the verbal working memory analysis and the limited number of methamphetamine studies included in the contextual verbal and visual memory analyses, significant differences between groups of studies should be interpreted with caution.

There are several important considerations to be made for future directions based on findings from this study. First, it is important to acknowledge that diverse neural networks, as opposed to single neuroanatomical structures, subserve the cognitive functions noted in this study. Therefore, future research may benefit from adopting a network-based approach to assess potential differences in functional connectivity between the two drugs. Second, given that the majority of studies examining cognitive remediation approaches have involved subjects with acquired brain injury, the discussion of their efficacy and possible application to cocaine and methamphetamine users requires empirical evaluation. Finally, although the subgroup analyses examining differences between crack versus cocaine and methamphetamine versus amphetamine did not reveal differences, these analyses were underpowered due to the limited number of studies available. The field of stimulant research would benefit from direct comparisons between crack and cocaine users, as well as amphetamine and methamphetamine users.

Finally, several unique cognitive weaknesses have been linked with the use of cocaine versus methamphetamine. Such distinctions may result in different behavior patterns that could impact treatment planning and triggers for relapse. Treatment success rates are similar between cocaine and methamphetamine users (152,153), however, regardless of the drug, a noteworthy percentage return to frequent use by the first posttreatment year (154,155). Individuals with compromised neuropsychological functioning are less likely to benefit from traditional treatments since cognitive difficulties limit their ability to engage in change (128). Tailoring behavioral treatment plans based on the unique neurocognitive difficulties associated with cocaine and methamphetamine use may aid in reducing relapse rates.

Declaration of interest

The authors report no relevant financial conflicts.

Supplemental data

Supplemental data for this article can be accessed on the publisher’s website.

Funding

This research was supported by NIDA award DA 026306 (Dr. Igor Grant) and seed grant from the California School of Professional Psychology (Dr. Alexander O. Hauson).

Additional information

Funding

This research was supported by NIDA award DA 026306 (Dr. Igor Grant) and seed grant from the California School of Professional Psychology (Dr. Omar Alhassoon).