Subjects

Patients aged 18-65 years with a DSM—IV (American Psychiatric Association, 1994) confirmed diagnosis of major depressive disorder (single episode or recurrent) were recruited from general practice clinics in the Tyne and Wear region. Patients had been entirely psychotropic-medication-free for at least 6 weeks before recruitment and were excluded if currently taking other medication active in the central nervous system (CNS-active), including β -blockers or St John's wort, or if there was a comorbid medical/psychiatric diagnosis, or a recent history of alcohol/substance misuse. Severity of depression was assessed using the Montgomery—Åsberg Depression Rating Scale (MADRS; Reference Montgomery and ÅsbergMontgomery & Åsberg, 1979), the 17-item Hamilton Rating Scale for Depression (HRSD₁₇; Reference HamiltonHamilton, 1960) and the Beck Depression Inventory (BDI; Reference Beck, Ward and MendelsonBeck et al, 1961). DSM—IV criteria for melancholia were noted and the Newcastle Scale (Reference Carney, Roth and GarsideCarney et al, 1965) completed. The Modified Mini-Mental State Examination (Reference Teng and ChuiTeng & Chui, 1987) was administered to screen for dementia.

The control group consisted of subjects who were psychologically and physically fit (verified by examination) and had no recent history of illicit drug use or alcohol misuse. Controls were excluded if they had a history of psychiatric illness (personally or in a first-degree relative) or a BDI score > 7. Current alcohol intake was less than 28 units/week for males and 21 units/week for females.

Patients and controls were matched for age, gender, premorbid IQ (National Adult Reading Test (NART); Reference NelsonNelson, 1982), years of formal education and season of testing. Females were matched for phase of menstrual cycle (see Table 1). All subjects had English as a first language. Subjects were tested as soon after recruitment as possible to minimise delay in treatment and in all cases treatment was commenced within 1 week of the initial assessment. The study was approved by the Newcastle and North Tyneside Health Authority Joint Ethics Committee and all subjects gave written informed consent.

Table 1

Demographic details for subjects with depression and for controls

	Depression			Controls
	Mean	s.d.	Range	Mean	s.d.	Range
Age (years)	32.9	10.6	19-61	32.3	11.4	18-55
NART	107.9	10.7	85-123	109.6	7.6	89-122
Formal education (years)	13.3	2.4	10-16	14.2	1.7	11-16
Gender (male:female)		15:29			15:29
Season1		9:11:16:8			7:11:19:7
Menstrual cycle2		13:7:5 (3 unknown)			18:8:3

Neuropsychological testing

Subjects were tested at 14.00 h, testing taking approximately 90 min to complete. The battery was designed to assess a broad range of cognitive domains. Pen-and-paper tasks were administered according to standardised instructions (Reference LezakLezak, 1995) and computerised tests from the Cambridge Neuropsychological Test Automated Battery (CANTAB) according to CANTAB manual protocols, on a PC fitted with a colour touch-screen monitor. These tests are briefly summarised below; detailed descriptions are available elsewhere (Reference Owen, Sahakian and SempleOwen et al, 1995; Reference Young, Sahakian and RobbinsYoung et al, 1999).

Psychomotor performance tests

Learning and memory (verbal) tests

Learning and memory (visuospatial) tests

Sustained attention and executive function

Statistical analysis

Demographic and neurocognitive data were assessed by analysis of variance (ANOVA) with group (depression or control) as the between-subject factor. Where tests had more than one level, an additional within-subjects factor of ‘time’ or ‘problem level’ was added and analysed by repeated-measures ANOVA, within a multivariate general linear model. Degrees of freedom were adjusted using the Huynh—Feldt epsilon if the assumption of sphericity was violated. For clarity, unadjusted degrees of freedom are reported. To reduce skewness, measures of response latency were logarithmically transformed (base 10) and errors on the spatial working memory task were square-root transformed. Where data could not be transformed, non-parametric tests were employed. Estimates of effect size were calculated for untransformed data using the formula (μ_MDD — μ _controls)/σ_pooled, where MDD denotes major depressive disorder (Reference HowellHowell, 1999). Statistical analyses were carried out using SPSS version 9 (SPSS, 1998).

Psychomotor performance tests

Learning and memory (verbal) tests

Learning and memory (visuospatial) tests

Simultaneous/delayed matching to sample

Accuracy on the simultaneous trial did not differ between groups (F=0.30, d.f.=1,78, P=0.59). For delayed trials, the group with depression performed significantly worse than controls (F=5.79, d.f.=1,78, P=0.02). There was also a main effect of delay (F=7.55, d.f.=2,156, P=0.003), performance being worse at the 12-s delay (see Fig. 1). No effect of group was found on latency measures (F=0.34, d.f.=1,78, P=0.85).

Fig. 1

Percentage of correct responses for simultaneous and delayed matching to sample (mean±s.e.m.): MDD, major depressive disorder; SIM, simultaneous trial.

Sustained attention and executive function tests

Vigil continuous performance test

Patients with depression made significantly more errors of omission (U=694.5, P=0.04) and commission (U=684.5, P=0.04) than controls. In patients, there was no change in error rate over time; however, controls made fewer errors after the first 2 minutes of the test (χ² _F=12.6, d.f.=3, P=0.005; see Fig. 2). There was no difference between groups in response latency (F=1.04, d.f.=1,84, P=0.31).

Fig. 2

Mean (untransformed) Vigil omission and commission errors across each 2-min time quarter of the test: MDD, major depressive disorder.

Spatial working memory

There were significant effects of group (F=8.62, d.f.=1,85, P=0.004), level (F=284.0, d.f.=2,170, P<0.0005) and a group by level interaction (F=4.49, d.f.=2,170, P=0.013) in the number of between-search errors. Patients with depression made more errors than controls at the six- (t=2.91, d.f.=85, P=0.005) and eight-shape (t=2.92, d.f.=85, P=0.004) problems (see Fig. 3). The strategy score was also greater in patients with depression, indicating a less-efficient search strategy (F=8.22, d.f.=1,85, P=0.005).

Fig. 3

Spatial working memory between search errors (mean±s.e.m.): MDD, major depressive disorder.

The correlation between strategy score and total between-search errors revealed a linear relationship between these variables in both the depression (r=0.85, R²=0.72, P<0.001) and the control groups (r=0.60, R²=0.36, P<0.001).

Exploratory data analyses

Learning and memory

Verbal declarative memory as assessed by the RAVLT was preserved, with no differences between patients with depression and controls on immediate word span, learning or long-term recall and recognition. Although recall of the distractor list differed between groups, there was no evidence of proactive inhibition. Several earlier studies have shown that patients with depression were impaired particularly on tests of verbal learning and memory (e.g. Reference Austin, Mitchell and WilhelmAustin et al, 1999). However, such tasks are sensitive to the effects of some antidepressants (Reference Schmitt, Kruizinga and RiedelSchmitt et al, 2001) and the medication-free status of patients in this study could explain the discrepancy. One study examining patients who had been drug-free for 2 weeks showed no difference between patients with non-psychotic major depressive disorder and controls in verbal memory, whereas executive performance (Stroop) was impaired (Reference Schatzberg, Posener and De BattistaSchatzberg et al, 2000). Alternatively, the absence of impairment could be attributable to other factors such as the small proportion of patients with severe depression or melancholia in our sample. Austin et al (Reference Austin, Mitchell and Wilhelm1999), using Clinical Outcomes in Routine Evaluation and Newcastle definitions of melancholia, found that pronounced deficits in verbal learning, recall and recognition were observed in patients with melancholia compared with controls, whereas performance was preserved in patients without melancholia.

In our study, recognition memory for visually presented patterns and spatial locations was impaired in the depression group, with reaction times also increased in the former. Previously, such impairments have been reported in middle-aged (Reference Elliott, Sahakian and McKayElliott et al, 1996) and elderly patients with depression (Reference Beats, Sahakian and LevyBeats et al, 1996), although deficits are observed rarely in studies of younger in-patient and out-patient populations (Reference Grant, Thase and SweeneyGrant et al, 2001).

The impairment in accuracy found on the delayed matching-to-sample task has been reported more consistently in in-patients and out-patients across all age groups (Reference Beats, Sahakian and LevyBeats et al, 1996; Reference Elliott, Sahakian and McKayElliott et al, 1996). An exception to this is the study by Grant et al (Reference Grant, Thase and Sweeney2001), in which there was a general absence of neurocognitive impairment across a broad range of tests, although on some measures subjects with depression out-performed controls, being faster (delayed matching-to-sample latency and reaction time) and more accurate (selected tests of learning/memory). This is likely to be attributable to the very mild severity of depression in their patient sample.

Attention and executive function

Some of the most robust effects in the present study were found on tests of attention and executive function. This is in accordance with the findings of many previous studies (Reference Rogers, Bradshaw and PantelisRogers et al, 1998). Patients with depression made more errors of omission and commission on the Vigil continuous performance test. These were evident in the first quarter of the test and did not increase over time, suggesting that some of the impairments found in other tests could be the result of a general reduction in attentional resources or focus. As both types of error were greater in the patient group, it is not the case that patients with depression adopted a more conservative response style.

Although there were impairments on the majority of executive tests, there was no significant difference between patients and controls on the Tower of London task. This is perhaps unsurprising, given the multiplicity of processes subserved by the central executive, alongside evidence that even ‘dysexecutive’ patients can show intact performance on some putative measures of executive function (Reference Baddeley, Delia Sala and PapagnoBaddeley et al, 1997).

Although large effect sizes were present in tests of executive function, performance did not correlate with severity of depression, whereas selected performance (accuracy) indices of most tests of learning and memory did. It is possible that executive functions are more sensitive, being affected adversely irrespective of the severity of depression. The correlation between verbal and visuospatial learning and memory, however, suggests that such functions are relatively intact in milder depressive episodes, becoming more pronounced as severity increases. Alternatively, executive deficits could represent a relatively stable trait marker, whereas mnemonic impairment is related to clinical state. Indeed, some studies have failed to demonstrate any residual memory impairment after clinical recovery, whereas others show persistent impairment, particularly in aspects of executive functioning (Reference Beats, Sahakian and LevyBeats et al, 1996; Reference Paradiso, Lamberty and GarveyParadiso et al, 1997).

Factors affecting neurocognitive performance

The patients recruited in this study were younger (<65 years) adult out-patients with a moderate severity of illness and were drug-naïve or unmedicated for a sufficient period of time up to and including the neurocognitive test day to exclude the effects of psychotropic medication as a potential confounder. The patient sample was also clinically homogeneous, with only five patients fulfilling DSM—IV criteria for melancholia, none having a comorbid medical or psychiatric diagnosis, and the majority (68%) experiencing their first depressive episode. As described previously, all these factors affect the pattern and magnitude of the observed neurocognitive deficits.

Previous studies have demonstrated deficits in neurocognitive performance in patients both with and without melancholia, generally more pronounced in the former (Reference Austin, Mitchell and WilhelmAustin et al, 1999). The impact of depressive subtype is, however, complicated by a difficulty in differentiating between the effects of severity and subtype (Reference Austin, Mitchell and WilhelmAustin et al, 1999). Psychomotor slowing has been suggested as a feature of melancholic depression. It is found more commonly in older subjects with depression (Reference Beats, Sahakian and LevyBeats et al, 1996). As the subjects with depression in our study were neither elderly, nor on the whole had melancholia, it is unsurprising that we found no evidence of motor/psychomotor retardation as indexed by reaction times and performance on the DSST.

The effects of antidepressant and other psychotropic medication and their withdrawal could be of particular importance in studies of cognitive function in samples with less-severe depression or non-melancholic depression, where impairment could be more subtle. In general it appears that antidepressants with anticholinergic properties impair aspects of cognitive function, whereas there could be less effect of selective serotonin reuptake inhibitors (SSRIs) or monoamine oxidase inhibitors. However, a recent study suggests adverse effects of the SSRI paroxetine on verbal learning (Reference Schmitt, Kruizinga and RiedelSchmitt et al, 2001), probably related to anticholinergic effects, whereas sertraline was found to have positive effects on verbal fluency, related to dopaminergic effects. Some studies have attempted to control for antidepressant effects by using a wash-out period. Unfortunately, it could be that in some cases this has not been long enough as, although onset of antidepressant discontinuation symptoms is usually within 5 days of cessation, symptoms can last 1-3 weeks (Reference HaddadHaddad, 1998).

Current hypotheses on the nature of neurocognitive impairment in depression

To explain the cognitive impairment seen in depression, some authors have suggested that effort is the principal mediating factor, whereby reduced attentional capacity causes impairment on tasks requiring effortful processing (Reference Hasher and ZacksHasher & Zacks, 1979). There is some support for this hypothesis, although it has been argued that subjects with depression might be able to mobilise resources to complete effortful tasks unless decision-making is required (Reference Thomas, Goudemand and RousseauxThomas et al, 1999), and therefore that effort and attention are not necessarily synonymous. A recent meta-analysis demonstrated minimal effect sizes on tests of primary, semantic and working memory, with the most prominent effects being in the effortful encoding of information and an accompanying inefficiency of retrieving poorly encoded information from declarative memory (Reference Zakzanis, Leach and KaplanZakzanis et al, 1998). Our results do not support this hypothesis. Performance deficits were observed on tasks of visual and spatial recognition memory in the face of intact declarative verbal learning and memory. Recognition is assumed to be less effortful than recall; therefore, the pattern of results does not fit the profile that would be predicted from this hypothesis. It has also been argued that the neurocognitive deficits could simply be the result of reduced motivation to testing; however, again, the results do not support this. Deficits were not observed on all tests and did not become more pronounced towards the end of the test session, when it might be expected that motivation would be affected adversely. Furthermore, it has been argued that because of the relationship between affect and drive (or motivation), to study reduced motivation is, in some respects, to study depression itself (Reference Austin, Mitchell and GoodwinAustin et al, 2001).

Elevated levels of corticosteroids have been shown to impair learning and memory in humans. This has been demonstrated by acute and subchronic (Reference Young, Sahakian and RobbinsYoung et al, 1999) administration of exogenous corticosteroids in healthy volunteers and in conditions associated with a chronic elevation of endogenous cortisol levels, for example Cushing's disease (Reference Starkman, Giordani and BerentStarkman et al, 2001). Hypercortisolaemia, which is frequently described in major depressive disorder, has therefore been suggested to be one of the principal causes of neurocognitive impairment. In earlier work (Reference Young, Gallagher and PorterYoung et al, 2002), we found evidence of an elevation in cortisol/DHEA ratios in the patients with depression, which provides a more accurate estimate of ‘functional’ hypercortisolaemia, although no correlation was found with neurocognitive test performance in general. Work has suggested that optimum neurocognitive functioning may be dependent upon the relative occupancy of mineralocorticoid and glucocorticoid receptors in the brain (Reference de Kloet, Oitzl and Joelsde Kloet et al, 1999). Therefore, although measures could reveal state-related increases in hypothalamic—pituitary—adrenal axis activity, they do not assess more-subtle facets of dysfunction at the receptor level, such as glucocorticoid receptor-mediated negative feedback mechanisms. For this, moresensitive ‘activated’ tests such as the dexamethasone/corticotropin releasing hormone challenge could be required.

CLINICAL IMPLICATIONS

1. ▪ Patients with major depressive disorder show pronounced impairment in executive function, with additional effects on mnemonic function dependent upon severity of depression.

2. ▪ The neurocognitive impairment was found in a group of young, predominantly first-episode, out-patients without melancholia and was not because of the effects of psychotropic medication.

3. ▪ Neurocognitive impairment is an objective measure that may be used as a tool to investigate the abnormalities in brain function underlying major depressive disorder.

LIMITATIONS

1. ▪ The patients studied were predominantly not severely depressed.

2. ▪ The specific cognitive architecture or neural basis of the impairment was not fully demonstrated.

3. ▪ The temporal trajectory of impairment, in particular, the degree of recovery with symptom resolution, remains to be fully determined.