Stroke and episodic memory disorders

Abstract

Memory impairments are common after stroke, and the anatomical basis for impairments may be quite variable. To determine the range of stroke-related memory impairment, we identified all case reports and group studies through the Medline database and the Science Citation Index. There is no hypothesis about memory that is unique to stroke, but there are several important facets of memory impairment after stroke: (1) Every node of the limbic system implicated in memory may be damaged by stroke but very rarely in isolation and the combination of amnesia with the associated deficits often illuminates additional aspects of memory functions. (2) Stroke produces amnesia by damage to critical convergence white matter connections of the limbic system, and stroke is the only etiology of amnesia that can delineate the entire pathway of memory and critical convergence points. (3) Stroke also impairs memory, without causing classical amnesia, by damaging brain regions responsible for cognitive processes, some modality specific and some more generally strategic, that are essential for normal learning and recall.

Introduction

The fundamental organization of the limbic structures essential for normal learning and recall has been known for decades. Anatomical studies in nonhuman primates and rodents (Aggleton et al., 1987, Aggleton and Mishkin, 1984, De Olmos, 1990, Jolkkonen et al., 2002, Nauta, 1961, Roberts et al., 2007, Rosene and Van Hoesen, 1987, Russchen et al., 1985, Swanson, 1978, Swanson and Cowan, 1975, Swanson and Cowan, 1977, Zola-Morgan et al., 1982) have shown that these limbic structures form two related anatomical units. The first unit includes the hippocampal formation with projections to the septal region and the mammillary bodies of the hypothalamus (fornix), then from the mammillary bodies to the anterior nuclei of the thalamus (the mammillothalamic tract), followed by projections to the cingulate and retrosplenial cortices (anterior limb of the internal capsule), and finally returning to the hippocampus (cingulum)—the so called Papez circuit (Papez, 1995). The second unit involves the amygdala and projections to the septal nuclei and dorsomedial nucleus of the thalamus (ventral amygdalofugal pathway), and from the thalamus to the prefrontal cortex (thalamocortical projections) and returning to the amygdala via the uncinate fasciculus (see Fig. 1).

Insight about the neuroanatomy of human memory has come from many sources, and this knowledge came without study of patients with stroke. The role of the hippocampus was demonstrated by epilepsy surgery (Scoville & Milner, 1957) and expanded and focused by anoxic brain injury (Zola-Morgan, Squire, & Amaral, 1986) and encephalitis (Cermak & O’Connor, 1983). A role of the diencephalon was demonstrated by thiamine deficiency particularly in alcoholics (Victor, Adams, & Collins, 1971) and extended by penetrating brain injury (Squire, Amaral, Zola-Morgan, Kritchevsky, & Press, 1989). These injuries have been considered relatively pure amnesia, but it is a very rare patient with any of these disorders or injuries who has purely amnesia. Only by compiling cases from disparate causes, with different associated deficits and with different lesion overlaps has core knowledge of the neuropsychology of amnesia emerged. Stroke has contributed to this knowledge. We will argue that a similar compilation of cases from diverse causes informs our understanding of the effects of non-limbic lesions on memory, and stroke has contributed to this understanding as well.

There are many reasons that stroke is often considered the optimal experiment of nature for unraveling the specific regional effects of brain disease on cognition. There are usually no pre-illness neurological issues to complicate interpretation of the effects of the index stroke (as in alcoholic Korsakoff's disease). With abrupt onset there is no concern about adaptive changes in the brain before the initial assessment (as in chronic epilepsy). Given time for accommodation and compensation of less specific acute effects of injury, post-acute impairments can be confidently attributed to the structural injury, and stroke generates relatively discrete lesion boundaries for mapping regional effects minimizing concerns about co-occurring diffuse, unmappable injuries (as in anoxia or encephalitis).

There are some limitations to reliance on stroke. Patients with stroke are often elderly and the effects of stroke need not be the same across the lifespan, nor can the effects of aging or undiagnosed degenerative diseases of aging always by eliminated. Whatever their age, patients with a clinical stroke often have accumulated years of subclinical vascular injury that can never be entirely discounted by imaging. Infarcts are constrained by the common patterns of vascular supply to the brain (see Fig. 2), and only a portion of the combinations of lesions that might be informative can be expected to emerge from a predictable vascular system. Hemorrhages are constrained by the locations of vessels that are predisposed to rupture and by planes of dissection. Although hemorrhages generate a greater range of lesion patterns, they are also associated with a greater amount of indirect, diffuse injury. Because neither respects functional anatomy, memory deficits from strokes frequently co-occur with language, visuospatial, and/or executive deficits. Finally, precise localization following a stroke is never incontrovertible; diaschisis can confuse lesion effects in the acute phase and plasticity and functional reorganization can obscure lesion effects in the later phases.

Between 20% and 50% of patients who survive a stroke complain about memory difficulties (Nys et al., 2005, Rasquin et al., 2002, Sorensen et al., 1982, Wade et al., 1986). There may be many specific causes, and not all of them directly related to the stroke: depression, medication effects, sleep disorders. We began our literature review with the assumption that there are two broad, directly stroke-related damage mechanisms. The medial location, complex connectivity and considerable bilateral integration of pathways make the limbic circuitry somewhat impervious to vascular injuries, but as a first mechanism, infarction or hemorrhage may involve any structure within the hippocampal or amygdala circuits (amnesia lesions) (Table 1). The second mechanism would be stroke outside of limbic circuitry impairing perceptual, cognitive, attentional or executive capacities essential for some aspect of memory (process lesions) (Table 2). Despite some limitations, stroke is the only etiology with the potential to inform and compare the effects of injury to every structure relevant to memory or learning, to demonstrate every critical pathway between those structures and to highlight the role non-limbic brain regions play in memory.