An Update on Postoperative Cognitive Dysfunction
Postoperative cognitive dysfunction or decline (POCD) is increasingly recognized as a common phenomenon after major surgery [1–4]. Because older age is a strong preoperative risk factor of POCD [2,5], the incidence of POCD is expected to increase as the population of older surgical patients grows. Improving the measurement of POCD and identifying its etiology is clinically important, as recent studies have associated POCD with impairments in daily functioning [6], premature departure from the labor market [7], and dependency on government economic assistance after hospital discharge [7].
Broadly speaking, POCD refers to problems in thinking and memory after surgery. POCD is not yet recognized in the International Classification of Diseases and is not listed as a diagnosis in the Diagnostic and Statistical Manual [8]. The term POCD is used mostly in literature to represent a decline in a variety of neuropsychological domains including memory, executive functioning, and speed of processing. POCD has been defined in a consensus statement as “a spectrum of postoperative central nervous system (CNS) dysfunction both acute and persistent … including brain death, stroke, subtle neurologic signs and neuropsychological impairment [9]. ”
POCD should be distinguished from delirium or dementia. Delirium describes an acute confusional state featuring disturbances in attention and decreased awareness of the environment [10,11]. Delirium symptoms fluctuate during the course of the day, and the patient often is disoriented. In addition, hallucinations and inappropriate communication or behavior may be observed in the presence of delirium. In contrast, a typical patient with POCD is oriented but exhibits significant declines from his or her own baseline level of performance on one or more neuropsychological domains [12–15]. After surgery, changes in cognitive status may present in the form of a frank delirium or POCD, or both. POCD differs from dementia, which describes a chronic, often insidious, decline in cognitive function. Alzheimer’s disease remains the most common form of dementia, but there is considerable overlap in neurodegenerative disease.
Selection of neurocognitive tests to measure POCD
Most studies of POCD have focused on describing changes in brain functioning, and more specifically, have studied either performance-based or self-report perceptions of changes in memory, executive function, attention, learning, language, visual spatial skills, mathematics, motor function, and anxiety or depression [16]. The selection of neurocognitive tests to document these cognitive changes varies extensively between studies [16]. This variability in choice of tests may be because of the absence of a clear, theoretically derived, and empirically tested model that describes the causes and outcomes of cognitive changes associated with the surgical experience. Therefore, while a consensus-recommended battery of tests was established more than 10 years ago [9], it was not accompanied by this necessary model [17], nor did it have an explanation of how the tests within the battery met the consensus definition of POCD. Without the adoption of such a model, it is difficult to conclude whether tools intended to detect POCD are, in fact, assessing the presence of POCD.
Since the consensus guidelines were published in 1995, diagnostic criteria for POCD have been suggested. One stated criterion specified that there must be significant changes in neuropsychological tests involving several of the following domains: learning and memory, attention, executive functioning, and language [18]. Additional areas for assessment suggested by Deiner and Silverstein include declines in perception and abstract thinking [19]. Although these guidelines provide guidance for the selection of tools to assess for POCD, they were not accompanied with specific justification for the choice of cognitive domains that should be assessed when testing a patient for POCD [17].
Recent reviews of POCD after cardiac [16] and noncardiac [20] surgery reveal that the 2 most commonly assessed cognitive domains assessed were (1) learning and memory and (2) attention and concentration. Of the studies included in these reviews, 97% of studies including patients undergoing cardiac surgery included a memory and learning test and 70% of studies including patients undergoing noncardiac surgery included a memory and learning test. Similarly, 94% of studies including patients undergoing cardiac surgery included an attention and concentration test and 57% of studies including patients undergoing noncardiac surgery included an attention and concentration test. About one-third of studies included tests of verbal and language skills and tests of visual and spatial skills. Still fewer studies included tests of numerical reasoning (6% for both types of surgical subjects) or executive function (14% for cardiac surgery and 6% for noncardiac surgery subjects). Few studies reported the percentage of subjects who experienced significant decline in performance after surgery for each cognitive test. Consequently, it is difficult to determine whether patterns of deficits in specific domains emerge across studies. One review concluded that declines in memory and attention and psychomotor function were consistently found after coronary artery bypass graft (CABG), but cautioned that the domains of POCD detected are likely to be a function of the tests used to assess POCD [21]. Increased reporting of results for individual tests would help inform the development of a conceptual framework that explains which domains of cognitive functioning are expected to be affected by different surgical experiences.
Many studies have used composite measures of cognitive functioning to assess patients for the presence of POCD. Nearly a third of studies of POCD in patients undergoing noncardiac surgery incorporated composite measures of functioning in their protocol [20]. The most commonly used composite measure was the Mini Mental State Examination (MMSE); it was included in 21% of reviewed studies. An important debate is whether composite measures and composite scores should be used to detect POCD. Composite measures and composite scores do not differentiate which areas of neuropsychological functioning are affected by the surgical experience. Understanding the specific areas of neuropsychological functioning involved in POCD would not only advance conceptual models of POCD, but would also help in specifying which self-care activities are likely to be affected by POCD. In contrast, Rasmussen and colleagues [22] suggest that the use of composite tests or scores is not problematic because the purpose of the postoperative neuropsychological testing is to detect the presence of general, rather than specific, changes in cognitive functioning.
Administration of neuropsychological tests
There are many procedural issues that should be considered when administering a neuropsychological protocol for the purpose of detecting POCD [22,23]. Those issues might be categorized into the following: test selection, test administration, and test scoring for the purpose of detection of POCD. Test selection should be guided by choosing tests that have validity for detecting change in functioning in those domains expected to be negatively affected by the surgical experience. In addition, there are several practical issues that should be considered when choosing tests for the assessment of POCD. For example, it is important to choose tests with difficulty levels that do not result in floor effects (many subjects scoring the lowest score possible) or ceiling effects (many subjects scoring the highest score possible) [23]. Tests that do not have floor or ceiling effects are likely to have greater sensitivity to detecting a change in functioning associated with surgery. Choosing tests with parallel versions reduces potential practice effects from remembering test stimuli from earlier administrations. This is particularly problematic for word-list learning tasks because some subjects are able to recall words from a prior administration of the task. The parallel versions should be administered in a different order between subjects (eg, ordered according to a Latin-square design) to avoid potential bias in estimates of change due to differential difficulty levels of different forms [24]. Tests should be validated for the language in which they will be administered. For example, the difficulty of word-list generation tasks that require the subject to list as many words as possible that begin with a specific letter varies depending on the letter specified [25]. Difficulty levels will vary across languages for the same letter.
Administration of tests used to detect POCD should be standardized across occasions and subjects. Consensus recommendations include that testing be conducted by “the same suitably qualified and trained individual and that the tests minimize subjectivity and be performed in a standardized manner” [9]. This recommendation is meant to reduce variance in subjects’ tests scores that cannot be ascribed to the subjects’ ability alone. Because preoperative performance may be negatively affected by surgery-associated anxiety [23], it has been suggested that the administration of mood and anxiety scales with the neuropsychological tests would allow for statistical adjustment of cognitive test scores by subjects’ mood state [22].
Although a variety of scoring methods for the detection of POCD have been used across studies, investigators generally agree that scoring methods should consider (1) baseline performance, (2) practice effects, and (3) change on more than one neuropsychological test [9,17,23,26]. Baseline assessments allow determination of whether an actual change in cognitive functioning occurred subsequent to the surgical event. Nearly half of studies of POCD have been conducted in adults undergoing cardiac surgery, a population at risk for cognitive changes because of underlying heart or vessel disease [27]. Practice effects refer to improvement in performance because of familiarity with test procedures and can occur in patients with [28] and without existing cognitive impairment [29]. A common way of measuring practice effects is to measure the average improvement in performance for a group of matched controls. An important question is how well the matched controls truly match the surgical population to which they are compared. For example, studies that provide comparisons between subjects and controls reveal that controls differ from subjects, with fewer males [5,30], lower depression levels [5], lower rates of comorbidity [31], and lower attrition rates in the control group [30].
Timing of assessment for POCD
Early assessments of POCD likely capture a different phenomenon than what late assessments of POCD capture, and each are accompanied by a unique set of issues. Surgery-related factors may affect test performance in the immediate postoperative period, including acute pain [32,34], the use of drugs [24,35], nausea, limited mobility, and fatigue. Thus, it has been argued that patients should not be evaluated for POCD until at least 1 week postoperatively [9,14,36]. Recent evidence suggested this delay might be arbitrary, as negative outcomes are associated with POCD detected in the first week after surgery. In a 2008 study of patients undergoing noncardiac surgery, POCD detected at hospital discharge (mean duration of stay, <7 days) was associated with an increased risk of death within the first 3 months after surgery [5]. Restricting testing to only the later postoperative period is also problematic because many patients are already discharged within 1 week after surgery. In fact, recent data from the National Center for Health Statistics showed that the average lengths of stay for patients between the ages of 45 and 84 years are currently between 5.0 and 5.6 days [37]. It has been reported that the average length of stay of patients older than 50 years who underwent major noncardiac surgery with no postoperative complications was 4 days [38]. In our study, 88% of those identified with POCD were discharged within a week of surgery [34]. In addition, Rohan and colleagues [39] reported that 47% of patients undergoing minor noncardiac surgery were found to have POCD before discharge. These findings suggest that limiting screening for POCD to 7 days after surgery could result in missed recognition of POCD in many surgical patients. In-hospital patient education usually includes detailed instructions for wound care, administration of new medications, symptom monitoring, and details of needed restrictions in daily activities. If postsurgical patients are experiencing POCD at the time they are given self-care discharge instructions, their ability to understand and recall these instructions may be limited and may put them at risk for postsurgical complications.
The rate of patient attrition is noted to be lower in studies assessing for early rather than late POCD. Newman and colleagues [20] reported a 5.4% attrition rate for evaluations performed between 7 and 21 days after noncardiac surgery; 19% for evaluations between 22 days and 132 days; and 17% for evaluations beyond 6 months postoperatively. Patient attrition may be selective. In cohort studies, patients unavailable for follow-up were more likely to be older patients who had worse baseline cognitive performance [40] or were sicker [41]. If patients experiencing cognitive decline are more likely to decline assessment, this selective attrition will bias study results toward the null, obscuring the detection of cognitive changes postoperatively.
POCD assessments that occur in the immediate postoperative period are important for elucidating the relationship between POCD and delirium. Because POCD and delirium both feature deficits in attention, whether they are related events on a continuum or distinct conditions remains unclear. In a retrospective analysis of the International Study for Postoperative Cognitive Dysfunction (ISPOCD) research data, patients with postoperative delirium had a higher incidence of POCD 1 week postoperatively [40]. The ISPOCD study was not initially designed to measure postoperative delirium as a primary outcome, so validated measures of delirium were not incorporated into the study protocol. It is not clear whether detection of delirium via chart reviews and the Mini Mental State Examination (MMSE) performance has similar sensitivity and specificity for detecting delirium as that of commonly used measures such as the Confusion Assessment Methods (CAM) [11]. Although the ISPOCD results seem to suggest that postoperative delirium and POCD appear to be discrete events, other research provides support to the continuum hypothesis, specifically, that POCD is a subclinical from of delirium. For example, Monk and colleagues [5] found that patients who were delirious after major noncardiac surgery were also more likely to have POCD at hospital discharge. Furthermore, areas of cognitive functioning that show decline in patients with POCD such as attention are common with criteria for detecting delirium. In a study of older orthopedic patients, Lowery and colleagues [42] identified a group of patients with “subsyndrome delirium,” defined as those not meeting the CAM criteria for delirium but showing a decline in global cognitive functioning as measured by the MMSE. These patients exhibited greater declines in performance on attention tasks than did nondelirious patients within the first week after surgery. Because most cases of delirium occur in the early postoperative period, an improved understanding of the relationship between POCD and delirium will be derived from additional studies that perform neurocognitive testing and delirium assessment simultaneously within the first several days after surgery.
Pathophysiology of POCD
The exact pathophysiology of POCD remains undefined. Previous studies of POCD have focused on investigating the risk factors associated with early POCD. Table 1 describes the variables that have been shown to be associated with early/intermediate POCD. In terms of patient-related baseline factors, or sometimes called predisposing factors, increasing age and lower levels of education have been identified as the main ones in the early study by the International Study on Postoperative Dysfunction (ISPOCD) [2]. In a subsequent study by Johnson and colleagues [3]