DISCUSSION

The main reason why previous studies have paid little attention to cognition in the early phase of stroke has been uncertainty about the reliability and predictive value of such an early examination. It has been argued that a number of stroke-related problems such as fatigue, fluctuating level of arousal, or emotional distress may exacerbate or obscure cognitive performance in this stage (Lezak et al., 2004). Nevertheless, the present study demonstrates that an early concise neuropsychological examination can be used to predict long-term cognitive performance as assessed with a more extensive examination in patients with a first-ever stroke. Therefore, such an examination could assist in a more appropriate allocation of rehabilitation resources. Moreover, interventions aimed at restoration and/or compensation of impaired cognitive functions could start off in an earlier stage, which might be important for the effectiveness of the treatment (Paolucci et al., 2000; Biernaskie et al., 2004). There were two exceptions with respect to the specificity of the predictions, however. First, an executive disorder rather than perceptual performance in the early phase of stroke predicted the presence of a long-term perceptual/constructional deficit. While there is some evidence that executive dysfunction is related to the severity of unilateral neglect after stroke (Manly et al., 2002; Rusconi et al., 2002), the present study shows that this also applies to visual perception and construction in general. Particularly, the constructional component in this domain may benefit from an intact planning capacity and the ability to use structure (Elderkin-Thompson et al., 2004). Second, abstract reasoning emerged as the best predictor of long-term language performance. In a similar vein, previous studies have shown a close association between abstract reasoning impairment and aphasia (De Renzi et al., 1966; Gainotti et al., 1986; Baldo et al., 2005).

The majority of patients who were cognitively unimpaired immediately after the stroke retained the same level of cognitive performance at follow-up as healthy controls, suggesting that there is no evidence of a generalized insidious cognitive deterioration in this stroke population, at least not in the first six months after the stroke. An important note here is that cognitive impairment was defined according to very stringent criteria (at least one domain score < −1.65 which is associated with the fifth percentile level of control performance). Therefore, some of these patients may have shown a more subtle cognitive decline compared to their premorbid level. For example, in a previous publication from our group, a small deterioration in cognitive performance has been demonstrated in patients with a single lacunar stroke (van Zandvoort et al., 2001). Although there was no generalized deterioration for the unimpaired group as a whole in the present study, a few patients developed new cognitive disorders as compared to baseline, which could be the consequence of increasing brain damage due to a growing amount of white matter lesions, cerebral atrophy, or silent infarcts (Vermeer et al., 2003). However, for the group as a whole, preexistent brain pathology was unrelated (1) to cognitive performance at baseline or (2) to the degree of cognitive recovery between baseline and follow-up.

Patients who were cognitively impaired in the early phase of stroke still performed worse than healthy controls at follow-up, especially if multiple cognitive disorders were present at baseline. Nevertheless, improvement as compared to baseline seemed to be the rule rather than the exception in all cognitive domains. Previous studies have reported global cognitive improvement rates ranging from 10% to 50% (Desmond et al., 1996; Tham et al., 2002; Ballard et al., 2003; Rasquin et al., 2004), depending primarily on the time interval between the stroke event and the neuropsychological assessments, and on the criteria used to define recovery. In the present study, domain-specific recovery occurred in 41% to 83% of cases, depending on the nature of the cognitive deficit. Recovery in the visual domains (i.e., in visual perception/construction and visual memory) was the most prevalent, whereas recovery in abstract reasoning and language was the least common. The prevalence of recovery in our study was much higher than in the aforementioned studies, and this is probably related to the fact that our baseline assessment occurred at a much earlier stage, leaving more room for improvement. Moreover, in contrast to previous studies (Desmond et al., 1996; Hochstenbach et al., 2003; Patel et al., 2003), we only included initially impaired patients in our estimation on the prevalence of cognitive recovery. Finally, we studied patients with a first-ever stroke without preexistent cognitive or functional problems, and this population probably has a larger potential for cognitive recovery than the stroke population in general (Jorgensen et al., 1997).

The high prevalence of domain-specific recovery in our study indicates that cognitive functions have a dynamic (adaptable) rather than a static (unchangeable) course after stroke. This finding is particularly important, as patients and their caregivers and some general practitioners often assume that cognitive decline cannot be restored following stroke because of the popular association with dementia. It has been argued that the mechanism underlying normal learning is the same mechanism that is implicated in the recovery of function following acquired brain damage, that is, the so-called Hebbian learning mechanism, which involves experience-dependent dendritic sprouting (Robertson & Murre, 1999). Moreover, there are some indications of neuronal regeneration following brain damage (Eriksson et al., 1998). The location of brain plasticity responsible for cognitive recovery is still unclear however, with evidence of both peri-lesional changes as well as contralateral reorganization (Calvert et al., 2000; Perani et al., 1993), possibly also depending on the interval post-stroke and on the nature of the cognitive deficit.

Several factors emerged as important independent predictors of domain-specific cognitive recovery, with some factors associated with recovery in multiple cognitive domains and some factors associated with recovery in isolated cognitive domains. A high level of preexistent ability (NART score) was an excellent predictor of a good recovery in almost all cognitive domains. One plausible reason is that these patients are more capable of creating a suitable compensation strategy to circumvent their cognitive deficit. Moreover, it has been suggested that factors such as intelligence or education provide the brain with a “cognitive reserve capacity” (Staff et al., 2004). The concept of cognitive reserve posits that individual differences in how tasks are processed might provide a differential reserve against brain pathology or age-related changes (Stern et al., 2003). The neurobiological correlates of this “cognitive reserve” are still uncertain, but animal studies have demonstrated that exposure to a more stimulating environment is associated with a greater synaptic density and more complex neuronal connections (Kolb, 1999). According to this view, a denser connectivity implies more opportunity for cognitive recovery (e.g., Kolb, 1999; Robertson & Murre, 1999). A younger age was also associated with a better recovery in executive functioning, language, and visual and verbal memory. According to the aforementioned view, age reduces the neuronal connectivity and might therefore result in a worse cognitive recovery after brain damage (Robertson & Murre, 1999). Gender did not influence recovery in any of the cognitive domains in our study.

Several lesion locations were related to recovery in cognitive functioning. A lesion involving the frontal lobe was related to poorer recovery in visual perception and construction. Occipital lesions compromised recovery in visual perception and visual memory, probably related to the presence of visual field defects in these patients. Temporal lesions compromised recovery in executive functioning and verbal memory. With respect to memory, the temporal cortex is believed to be responsible for the consolidation of memory traces (Miyashita, 2004), which may explain the association we found with a poorer recovery in verbal memory. In contrast to previous studies that have reported conflicting findings with evidence of both a left and a right hemisphere advantage related to cognitive recovery (Desmond et al., 1996; Hochstenbach et al., 2003), lesion side was not a predictor of recovery in the present study. Lesion volume emerged as an independent predictor of visual memory, but not with respect to recovery in other cognitive functions. These findings indicate that lesion location affects cognitive recovery more than lesion volume when considering the simultaneous influence of multiple lesion characteristics. Prior studies have indeed shown severe cognitive impairment in patients with only small but strategically located lesions (Auchus et al., 2002; Szirmai et al., 2002).

Diabetes mellitus was related to recovery in abstract reasoning, but other vascular risk factors were not related to cognitive recovery in any domain. The link between diabetes mellitus and poor recovery in post-stroke cognitive functioning (Desmond et al., 1996) and cognitive functioning in general (Biessels et al., 2002; Kanaya et al., 2004) has been demonstrated before. Diabetes mellitus has also been identified as a risk factor for vascular dementia (Biessels et al., 2002). In this respect, it has been shown that both pathological increases and decreases of glucose concentrations may affect cognition. Also, the neuroradiological alterations in patients with diabetes mimic those observed in the aging brain, and therefore it has been suggested that diabetes causes an acceleration of the aging process in the brain (Biessels et al., 2002).

Finally, neither stroke severity, as measured with the NIHSS, nor the presence of depressive symptoms at baseline mediated cognitive recovery. In contrast, the severity of unilateral neglect immediately after stroke was related to a poor recovery in visual memory, but this relation disappeared in the multivariate model. It has been hypothesized that unilateral neglect probably arises from a combination of two deficits, that is, a rightward bias plus a deficit in spatial working memory (i.e., retaining locations already searched) (Husain et al., 2001). Our findings support this view and suggest that even when the lateral bias disappears, the visual memory capacity may still be affected in the longer term.

Some limitations of this study should be addressed. First, we only had a single follow-up examination between six and ten months after stroke. Although the largest amount of spontaneous recovery will have taken place within this time range, it is possible that we underestimated the prevalence of recovery. Alternatively, the course of cognitive functioning might change after the first six months, as one recent study has shown that 22% of patients with a stroke or TIA demonstrated a decline in cognitive functioning between 3–6 months post-stroke and after one year (Sachdev et al., 2004). A potential underlying mechanism might be that the preexistent brain pathology that was initially silent continues to proceed and exacerbate cognitive performance in the longer term after stroke. Therefore, our cohort is in the process of a further follow-up to examine the nature of domain-specific cognitive change in the longer term. Second, as we were interested in associates of cognitive recovery exclusively related to the stroke, we included a selective stroke population without preexistent cognitive, neurological, or psychiatric deficits. Consequently, these findings cannot directly be generalized to the population with first-ever stroke as a whole, particularly given the fact that patients who were not included for follow-up demonstrated more cognitive impairments at baseline. In addition, 22 patients were too impaired in the early phase to be included in our study at all. This selection bias probably resulted in an underestimation of the frequency of impairment in the long term after stroke, and might have concealed predictors of cognitive recovery. In an attempt to avoid this selection bias as much as possible, we chose to administer a concise neuropsychological examination in the acute phase of stroke and to extend this examination at follow-up. It should be noted, however, that this approach might have influenced our findings with respect to the prevalence of domain-specific cognitive recovery in our sample. A final important remark is that we do not claim to have examined the degree and associates of spontaneous recovery. In this respect, it should be noted that cognitively impaired and unimpaired patients demonstrated a differential discharge destination and treatment pattern. Rather, we have examined outcome according to the status quo (standard practice) in the Netherlands, with cognitively impaired patients rightly receiving more treatment than cognitively unimpaired patients. It is sensible to assume that this probably affected outcome in cognitively impaired patients in a positive way. Moreover, cognitive rehabilitation in a certain cognitive domain (e.g., executive functioning) might be less effective than cognitive intervention with respect to another cognitive domain (e.g., language), and these differential effects might have influenced our findings to some extent. Future research in the effectiveness of cognitive rehabilitation is certainly warranted in order to disentangle treatment effects from spontaneous recovery.

In sum, identification and characterization of cognitive impairment in acute stroke provides valuable and specific prognostic information with respect to long-term cognitive functioning. The prognosis of acute higher-level visual disorders was the most favorable. Cognitive recovery after first-ever stroke was associated with younger age, higher preexistent ability, smaller lesion volume, lesion location, and diabetes mellitus.