Predictors of Cognitive Dysfunction after Major Noncardiac Surgery

After approval by the institutional review board, 1,064 patients undergoing major, noncardiac surgery at Shands Hospital at the University of Florida (Gainesville, Florida) gave their written informed consent and were enrolled in the study between February 1, 1999, and January 31, 2002. Inclusion criteria included adult patients who were aged 18 yr or older and scheduled to be admitted to the hospital as an inpatient for a minimum of 2 days after surgery. Additional inclusion criteria were surgery scheduled under general anesthesia that was expected to last 2 h or longer, fluency in English, ability to read, and the absence of serious hearing or vision impairment that would preclude neuropsychological testing. Patients were excluded if they were scheduled to undergo cardiac, carotid, or intracranial procedures or were not expected to be alive or available to complete testing at 3 months after surgery. Additional exclusion criteria included a score of 23 or less on the Mini-Mental State Examination (MMSE)8before surgery; a history of dementia or any disease of the central nervous system, including previous cerebral vascular accident with residual deficit; current use of tranquilizers or antidepressants; a current or past history of psychiatric illness; and alcoholism or drug dependence. Patients could not be included twice, even if they had an unrelated second procedure.

Previous studies have suggested that repeated neuropsychological testing may result in an improvement in test performance, or a practice effect.9,10To estimate the magnitude of this practice effect for the neuropsychological tests used in this study, we enrolled a group of 210 control subjects. These subjects were recruited from primary family members (spouses, siblings, or children) of patients enrolled in the study. With the exception of the surgically related items, inclusion and exclusion criteria for the patients and control subjects were identical. The control group underwent neuropsychological testing at the same time intervals as the patients. Control subjects did not undergo surgery or anesthesia during the study period.

The preoperative evaluation was performed within 14 days of surgery. Information about the patients’ demographic status, medical history, physical examination, education, and employment status were recorded. Current medications, alcohol use, smoking history, and previous surgical history were also recorded. The Charlson comorbidity score,11a weighted index that includes both the number and seriousness of comorbid medical conditions, was computed for all patients at hospital admission. The anesthesia provider also rated the patient’s presurgical comorbidity with the American Society of Anesthesiologists physical status.12 

All surgery was performed with general anesthesia, and there were no restrictions on the type of anesthesia or postoperative analgesia. All patients were visited daily during their hospital stay, and the medical records were reviewed for complications. At 1 yr after surgery, patients or their families were contacted via  telephone to determine their survival status.

A brief battery of neuropsychological tests was administered before surgery (within 14 days of surgery), on the day of hospital discharge (or at 1 week after surgery if still hospitalized), and 3 months after surgery. The neuropsychological assessment was performed in a quiet room with only the patient and the psychometrician present. If a patient had a second operation during the study period, he or she remained in the study, and the neuropsychological test results were included in the final analysis. The occurrence of a second operation in the first 3 months after surgery (before the 3-month neurocognitive testing) was captured as a variable and included in the analysis of factors predicting POCD.

Each examiner was trained on psychometric test administration and relevant interview techniques by the neuropsychologist (D.E.D.) involved in the protocol. All measures were administered and scored in a standardized manner to minimize differences between test administrators. Project investigators trained in neuropsychological assessment completed all data scoring and interpretation.

The neuropsychological test battery included the tests used to evaluate patients for POCD in the ISPOCD1 and 2 studies.3,13Three parallel forms of the tests were used in a sequence using a full Latin square design for tests other than the Stroop test. Patient assignment to one of six possible sequences was random. These tests primarily focus on memory and executive functions and evaluate the following:

1. Word learning: Visual Verbal Learning Test14based on the Rey’s Auditory Recall of words

2. Word recall: the number or words recalled from Visual Verbal Learning trials after a 20-min delay

3. Cognitive flexibility: Concept Shifting Test,15part C, based on the Trail Making Test

4. Distractibility: Stroop Color Word Interference Test16 

5. Working memory: Letter–Digit Coding17based on the Symbol Digit Substitution Test from the Wechsler Adult Intelligence Scale III, 3rd edition

Mood and pain assessment measures included the following:

1. Beck Depression Inventory18: This assessment screens for the presence and degree of depression in adults. It has been found to have high reliability and validity for identifying depressed patients across age groups, sexes, and cultures.

2. State Trait Anxiety Inventory19: This inventory has two separate scales for measuring state and trait anxiety. The state anxiety scale consists of 20 statements that evaluate how the respondent feels at the present time. The trait anxiety scale consists of 20 statements that assess how the respondent generally feels.

3. Numerical rating scale20for pain: 0 indicates no pain, and 10 indicates severe pain.

For the first 3 days after surgery, patients were visited by a member of the study team, and delirium was assessed using the confusion assessment method.21A patient was considered delirious if the confusion assessment method was positive on any one of these days. If a patient exhibited delirium at a testing time, the neuropsychological test battery was not performed.

During the preoperative visit and the 3-month postoperative testing visit, the patient’s functional status was estimated by the administration of the instrumental activities of daily living questionnaire,22which contained seven questions related to use of the telephone, shopping, domestic work, preparation of meals, and related activities. This questionnaire was completed by both the patient and an informant (a member of the patient’s immediate family). If a patient could do the activity without assistance, the question received a score of 0; if some assistance from other people was needed to do the activity, the score was 1; and if the patient could not do the activity, the score was 2 (scores ranged from 0 to 14).22Higher scores indicate increasing difficulty in engaging in daily activities.

The primary expected outcome of concern for the study was the occurrence of cognitive decline 3 months after surgery. To determine the presence of cognitive decline, we used the same definition of POCD that was used in the ISPOCD1 and ISPOCD2 studies.3,13To quantify the practice effect, we compared the average amount of change in performance for control subjects in each age group (18–39 yr, 40–59 yr, and ≥60 yr) for each measure between baseline and subsequent tests 1 week and 3 months later. For patients, we compared baseline (preoperative) scores with the 1-week and 3-month postoperative test results, subtracted the age-appropriate average practice effect from these scores, and divided the result by the age-appropriate control group SD to obtain a Z score for each individual test. This technique identifies patients with POCD by comparing the changes in test scores of an individual patient undergoing surgery with changes in the test scores of the age-matched control group over the same interval. The sign is adjusted so that positive Z scores indicate deterioration from the baseline test.3,10 

The Z score for all tests can be summarized as a composite Z score.3The composite score is calculated as the sum of the Z scores for an individual patient divided by the SD for the sum of the Z scores for the age-matched control group. A patient was classified as exhibiting POCD if the Z scores on two individual tests or the combined Z score was 1.96 or greater. This definition identifies patients with general deterioration in all tests or substantial deterioration on only some tests.

We used chi-square or Fisher exact test for bivariate analyses of categorical data (e.g. , testing association of POCD and age group) and nonparametric tests (i.e. , Wilcoxon signed-rank tests, Kruskal-Wallis, Spearman correlations) for analysis of numerical data. We modeled the binary outcome of POCD in a logistic regression to determine which demographic and perioperative factors were associated with cognitive decline at 3 months after surgery after adjusting for other factors. All factors that were significant in bivariate analyses were included in the logistic regression model. We used SAS software version 9.1 (SAS Institute Inc., Cary, NC) for all analyses.

The sample size in this study was determined for the purpose of analyzing the difference between POCD in elderly (60 yr or older) versus  middle-aged (40–59 yr) and younger (18–39 yr) patients. We choose these age classifications because the ISPOCD1 study evaluating POCD in the elderly included patients aged 60 yr or older and the ISPOCD2 study evaluating POCD in middle-aged patients included patients aged 40–59 yr.3,13The ISPOCD1 study3used Z score calculations and found that elderly patients had a 10% incidence of POCD at 3 months after major, noncardiac surgery. For the sample size calculation, we assumed that the incidence of cognitive decline at 3 months after surgery would be 5% for middle-aged patients and 3% for young patients. Using logistic regression to test for the presence of a linear trend that a patient would demonstrate a cognitive deficit at 3 months after surgery, we calculated that 375 patients per group would provide 80% power to see this difference based on a two-tailed significance level of 0.05. The sample size calculation was performed using the SIZ sample size and power module contained in the EGRET epidemiologic statistics software package (Cytel Software Corp., Cambridge, MA). This calculation determined that a sample size of 375 patients in each of the three age groups would allow us to estimate the incidence of POCD after 3 months in young, middle-aged, and elderly surgical populations with 2.9, 3.3, and 4.2% margins of error, respectively, when constructing 95% confidence intervals (CIs). We included approximately 70 control subjects in each age group who were not undergoing surgery so we could calculate the practice effect of repeated neuropsychological testing.

A total of 1,496 patients were assessed for eligibility in the study; 267 refused to participate and 165 did not meet the inclusion criteria. Therefore, a total of 1,064 patients were enrolled, with 331 (31%) patients in the young (18–39 yr), 378 (36%) in the middle-aged (40–59 yr), and 355 (33%) in the elderly (≥ 60 yr) group. Of the original 1,064 patients, 1,021 (96%) were available for testing at hospital discharge, and 926 (87%) were available for follow-up at 3 months after surgery. At the late test time, the dropout rate was evenly distributed among the age groups; 50 patients in the young, 42 in the middle-aged, and 46 in the elderly group. The reasons for loss to follow-up were death (26%), too ill to complete testing (6%), too depressed to complete testing (5%), patient refusal (18%), moved out of state (15%), and inability to contact the patients (30%). During the same time period, 210 control subjects were enrolled (74 young, 74 middle-aged, 62 elderly). Twenty-eight (13%) of the control subjects dropped out before the final testing period (10 young, 12 middle-aged, 6 elderly). Data from 49 of the patients and 5 control subjects in the middle-aged group were included in the ISPOCD2 article describing POCD in middle-aged patients.13 

The baseline characteristics of the patients included in the study are listed in table 1and were similar to those individuals who declined to participate in the study. There were more women in the young age group compared with the elderly group. Elderly patients had more comorbid conditions than younger patients, and this was a reflected in higher baseline American Society of Anesthesiologists physical status and Charlson comorbidity scores. Education levels (median educational level was 13 yr for all age groups) and preoperative pain scores were similar among the groups. Older patients had baseline MMSE scores, anxiety scores, and depression levels that were statistically lower than the other two groups. Characteristics of the surgical procedures and hospital stay are included in table 2.

Patients and their age-matched controls were similar with respect to age, sex, educational level, and preoperative MMSE and State Trait Anxiety Inventory scores (table 3). Young and middle-aged patients had significantly greater levels of depression on the Beck Depression Inventory than age-matched controls, whereas there was no difference in depression scores between elderly patients and their controls.

The mean scores for the neuropsychological tests at baseline and 3 months after surgery are presented in table 4. The incidence of POCD at hospital discharge and the 3-month test session are presented in figure 1and table 5. There were 367 patients with POCD at hospital discharge, and the incidence of cognitive dysfunction ranged from 30% to 41%. There were no statistically significant differences in POCD between the young group and either the middle-aged or elderly group; however, the elderly patients had a higher incidence of POCD than the middle-aged patients (P = 0.01). The prevalence of cognitive decline at hospital discharge was similar for control subjects of all age groups at this testing session (4.1% [95% CI, 0.9–11.5%] in the young, 2.8% [95% CI, 0.3–9.6%] in the middle-aged, and 5.1% [95% CI, 1.1–14.2%] in the elderly group). Cognitive impairment was significantly higher in age-matched patients compared with age-matched control subjects in all three groups at hospital discharge (P < 0.001; fig. 1). Because there was no statistical difference in the incidence of cognitive decline among the control groups at either testing time, the mean incidence for the control groups is illustrated in figure 1.

At 3 months after surgery, cognitive deficits occurred in 12.7% of elderly patients, which was significantly greater than the incidence in young (5.7%) or middle-aged patients (5.6%) (P = 0.001; fig. 1and table 5). The incidence of cognitive decline in the control subjects was similar for all age groups at the 3-month testing interval and was 6.3% (95% CI, 1.7–15.2%) in the young, 4.8% (95% CI, 1.0–13.5%) in the middle-aged, and 1.8% (95% CI, 0.0–9.6%) in the elderly group. At 3 months after surgery, when the prevalence of cognitive deficits in the young and middle-aged patient groups were compared with their age matched control groups, there were no significant differences. However, when the elderly patients’ cognitive abilities at 3 months after surgery were compared with their age-matched controls, there was a significantly higher prevalence of cognitive decline in this elderly patient group (P < 0.001; fig. 1).

Predictors of POCD at hospital discharge are shown in table 6. During hospitalization, 35 of the 1,064 patients (3.3%) enrolled in the study experienced delirium, with 4 (1.2%) young, 8 (2.1%) middle-aged, and 23 (6.5%) elderly patients (P = 0.003). Only 26 of these 35 patients (74.3%) were able to be tested at hospital discharge. Delirium during the hospital stay was associated with POCD at hospital discharge (P = 0.046), but not with POCD at 3 months after surgery (tables 6 and 7).

Forty-six patients (4.3%) had a second operation during the study period, with 13 patients undergoing the second operation before the hospital discharge testing and 33 patients between hospital discharge and the 3-month testing time. A second operation before hospital discharge was not associated with POCD at hospital discharge. Similarly, a second operation at any time after the initial surgery was not associated with POCD at 3 months after surgery (table 7).

Forty-seven (63%) of the 74 patients with cognitive decline at 3 months after surgery also had POCD at hospital discharge, whereas the other 27 patients (36%) did not have cognitive decline at hospital discharge (table 8). Cognitive dysfunction at hospital discharge was an independent predictor of POCD at 3 months after surgery (P = 0.0001; table 7).

Independent predictors of cognitive decline at 3 months after surgery are included in table 7. Of the significant bivariate predictive factors for POCD at 3 months after surgery, increasing age, history of cerebral vascular accident with no residual impairment, years of education, and POCD at hospital discharge remained significant in the multiple logistic regression analysis.

At 1 yr after surgery, we were able to contact 914 of the 926 patients (99%) (or their families) who had completed neuropsychological testing at 3 months after surgery. A total of 58 patients (5.5%) died within 1 yr of surgery, with 16 dying during the first week, 20 dying in the 1-week to 3-month testing interval, and 22 dying after the 3-month testing session. There were 8 deaths in the young, 16 deaths in the middle-aged, and 34 deaths in the elderly group. Patients with POCD at hospital discharge were more likely to die before the 3-month testing time (6.5%) compared with patients without POCD (3.4%) (P = 0.02). Likewise, patients who exhibited POCD at both hospital discharge and 3 months after surgery were more likely to die in the first year after surgery (10.6%) compared with patients with no POCD (2.1%), cognitive decline only at hospital discharge (2.86%), or cognitive decline only at 3 months after surgery (3.9%) (P = 0.02; fig. 2).

This is the first prospective longitudinal study to simultaneously investigate postoperative cognitive changes after major noncardiac surgery in adult patients of all ages. Our findings indicate that 30–41% of all adult patients experience POCD at hospital discharge. Patients in all age groups have improvement in cognitive function by 3 months after surgery, but the prevalence of late POCD was significantly higher in the elderly than in the young or middle-aged patients. Our study confirms the findings of ISPOCD1, which found that advancing age and lower educational levels are risk factors for the development of cognitive decline after noncardiac surgery.3In addition, we found that asymptomatic patients with a history of stroke and those with POCD at hospital discharge had a higher incidence of late (3 months after surgery) POCD.

There were baseline differences among the groups, but advancing age is responsible for the majority of these differences (table 1). Depression and anxiety were less common in elderly patients. These finding are consistent with previous research, which indicates that younger individuals in a community population experience a higher prevalence of depression.23Despite the statistical differences in depression scores, the mean values for all groups were well below the cutoff of 14, which is needed to diagnose depression with the Beck Depression Inventory.18Normative samples of young, middle-aged, and older adults have also found that mean anxiety scores were similar in the two younger groups and somewhat higher than those of the elderly group.19As expected, elderly patients had more comorbid conditions than patients in the young and middle-aged groups, and this was reflected in higher baseline American Society of Anesthesiologists physical status and Charlson comorbidity scores. Older patients had baseline MMSE scores that were statistically lower than the other two groups. However, these differences were minimal and not likely to be clinically significant.24 

The high rates of POCD (30–41%) at hospital discharge in our study were associated with multiple factors (table 6) and are in agreement with the high rates of early POCD found in the ISPOCD 1 and 2 studies.3,13Older patients with less education, higher American Society of Anesthesiologists physical status, more complicated surgery, and longer hospital stays were also more likely to exhibit POCD at hospital discharge. Age has repeatedly been identified as a risk factor and predictor of early POCD after cardiac and noncardiac surgeries.1,3The ISPOCD1 study also found that lower educational level predicted cognitive decline at hospital discharge after noncardiac surgery.3 

It was surprising that early POCD was significantly higher in elderly versus  middle-aged patients, but there was no statistical difference between elderly and young patients (table 5and fig. 1). The majority of the younger patients were tested 1 day earlier than the other two groups (median, 3 days after surgery compared with 4 days for the middle-aged and elderly groups; table 5). It is possible that younger patients had a higher incidence of early POCD than middle-aged patients because they had less time to recover from the effects of anesthesia and surgery before testing.

Patients with POCD at hospital discharge had a higher rate of opioid analgesic use than those with no cognitive impairment. Opioids have sedative effects that likely have a detrimental influence on patients’ performance on neuropsychological tests. Alternatively, patients still receiving opioids at hospital discharge may have been experiencing significant pain that would also detract from their test performance.25,26 

Our study is one of the few investigations of postoperative cognitive function that includes younger patients. We found a low incidence of delirium in all age groups in our study, but still demonstrated that elderly patients are much more likely to develop postoperative delirium than younger patients. Our study only included patients with a preoperative MMSE score of 24 or greater, and the relatively low incidence of postoperative delirium in our elderly patients is similar to that previously reported for elders with preoperative MMSE scores of 25 or greater undergoing hip surgery.27Investigations reporting higher incidences of delirium often include patients with lower preoperative MMSE scores.27 

Patients in the current study who developed postoperative delirium were more likely to have POCD when tested at hospital discharge but not at 3 months. It is unclear whether this relation represents a continuum of disease between delirium and POCD or whether the existence of the former simply confounds the sensitive neurocognitive tests used to diagnose POCD. None of the patients were delirious when tested; however, it is possible that some of the patients had subclinical delirium not detected by the confusion assessment method at the time of testing. Delirious patients have difficulty concentrating, distractibility, and a diminished attention span and have been shown to score lower on neurocognitive tests.28 

There are other potential etiologic factors of early POCD that we did not examine and are unable to confirm with this study. Disturbed cortisol secretion29,30and the release of proinflammatory mediators,31both of which are a part of the physiologic stress response to major surgery, have been implicated in cognitive decline in humans. Lingering anesthetic effects in the form of central anticholinergic or anticatecholeminergic32,33activity and sleep deprivation associated with hospitalization34may also have had deleterious effects on our patients’ neurocognitive test performance at hospital discharge.

The multiple factors that affect cognitive testing at hospital discharge make it difficult to determine whether patients who decline on neurocognitive tests at this time point have POCD or whether the declines are secondary to other factors. Hospital discharge may be too early to test for cognitive changes and postoperative testing should be delayed until patients have recovered from the effects of sleep deprivation, anesthetics, and opioid agents.

Cognitive performance improved for impaired patients in all age groups after hospital discharge, but approximately 13% in the elderly group demonstrated cognitive dysfunction at 3 months after surgery. This is in agreement with the findings of ISPOCD1, which reported that 10% of patients 60 yr or older had prolonged cognitive dysfunction 3 months after surgery.3Approximately 6% of middle-aged patients in our study exhibited POCD at 3 months after surgery, which is similar to the prevalence found in middle-aged patients at the same time point in the ISPOCD2 study.13Multiple logistic regression analysis of prolonged POCD identified four independent factors: advanced age, lower educational level, previous stroke with no residual impairment, and POCD at hospital discharge.

Our study provides further evidence that the elderly are at particular risk of prolonged POCD. Although it is well known that there is a gradual decline in cognitive function during normal aging, some elders seem to be at greater risk than others for the development of dementia35and progress rapidly to a diagnosis of Alzheimer disease.36Our elderly group could have included subjects with mild preexisting cognitive deficits that were not detected by the screening MMSE. It is possible that perioperative events may have hastened or unmasked ongoing cognitive decline in some of these patients.

We found that lower educational level was associated with prolonged cognitive decline as previously reported by the ISPOCD1 investigators and others.1,3,37Previous studies have shown that a larger cognitive reserve capacity is protective in the development of dementia in the elderly38,39and is associated with better functional status after traumatic brain injury40and cerebral infarction.41Our results are consistent with a study in a normal aging population that demonstrated protective effects of higher educational level or better intellectual endowment against the development of cognitive decline.42 

We were unable to demonstrate a relation between baseline cognitive function as measured by the MMSE and the development of POCD at 3 months after surgery (table 7). The MMSE is fundamentally a screening test, and it may not be sensitive enough to detect subtle degrees of neurocognitive impairment. The authors of a widely used compendium of neuropsychological tests state that the “presence and nature of cognitive impairment should not be diagnosed on the basis of MMSE alone.”43Future studies should determine whether more sensitive neuropsychological tests can be used to identify subtle cognitive impairment before surgery.

Patients with lower instrumental activities of daily living scores before surgery (i.e. , needed more assistance with activities before surgery) were more likely to develop POCD at 3 months after surgery (P = 0.0033; table 7). Our investigation did not demonstrate a correlation between a decline in instrumental activities of daily living scores and the occurrence of POCD at 3 months after surgery. This finding is in contrast to the results of the ISPOCD1 study, which demonstrated a significant correlation between declines in instrumental activities of daily living scores and cognitive dysfunction at 3 months after surgery.3Because 136 patients in our study were not available for testing at this time point (51 of these patients [36%] were unavailable secondary to death, severe depression, or illness), it is possible that we excluded the patients who had the greatest decline in functional status. However, a companion article evaluating the subtypes of cognitive impairment in the elderly patients in this study found that impairment in instrumental activities of daily living scores were related to the type of postoperative cognitive impairment.7We did, however, find a significant correlation between POCD and mortality which is of even greater clinical significance.

The finding that asymptomatic patients with a history of stroke had an increased incidence of prolonged POCD may also be related to the concept of cognitive reserve. Despite having no residual neurologic deficits at the time of surgery, these patients may have lost critical neural mass with their stroke, leading to a decrease in their cognitive reserve and an increased susceptibility to POCD. Preexisting brain dysfunction is a well-known risk factor for postoperative delirium in the elderly.44Our results support the conclusions of two previous studies that described the role of preexisting brain injury/dysfunction as a risk factor for POCD. Goto et al.  45reported that patients about to undergo coronary artery bypass grafting who had multiple cerebral infarctions on magnetic resonance imaging had an incidence of early postoperative cognitive decline three times that of patients with no infarctions. In this study, nearly two thirds of patients with preoperative magnetic resonance imaging findings of stroke were asymptomatic.45In a study of 140 elderly patients undergoing noncardiac surgery, French investigators found that preoperative cognitive impairment was a risk factor for prolonged POCD.37 

Patients experiencing cognitive decline at hospital discharge were at higher risk for late POCD when tested at 3 months postoperatively. Of the patients with POCD at 3 months, 65% were considered to have persistent POCD from their hospital discharge testing period. Two previous studies have demonstrated an association between POCD at 1 week and very long-term POCD (1–5 yr) after both cardiac1and noncardiac surgery.46The existence of persistent POCD suggests a perioperative injury to the brain resulting in prolonged problems with cognition. The nature of such an injury can only be speculated upon but could involve some of the proposed mechanisms for early POCD alone or in combination with other factors (i.e. , paradoxical embolization during orthopedic surgery or an underlying genetic or neurologic predisposition to brain injury).

It is unclear why some patients developed POCD between the assessment at hospital discharge and 3 months (table 8). One possible explanation for this delayed POCD might be that these patients experienced a progression of their underlying medical comorbidities or that they acquired new medical problems in the interval between testing sessions. Likewise, patients who developed clinical depression after their surgery might perform more poorly on the 3-month postoperative neuropsychological test battery. Stockton et al.  47reported that elderly patients who developed depression or new onset medical illness during the 12 months after surgery were more likely to manifest cognitive decline as the study progressed. The cause of cognitive decline in our delayed POCD group may be different than in patients with persistent POCD as reflected by significantly different 1-yr mortality rates in these two groups of patients. If the etiology of delayed POCD is not related to surgery, it might be more appropriate to refer to this problem as cognitive dysfunction and not delayed POCD.

This study is the first to report an association between the occurrence of POCD and increased mortality in the first year after major noncardiac surgery. Patients with POCD at hospital discharge were more likely to die before the 3-month testing time. Likewise, patients with POCD that persisted for the initial 3 months after surgery were at a significantly higher risk of death in the first postoperative year.

Our observation of an association between POCD and postoperative mortality is in agreement with numerous population-based studies reporting an increased risk of early mortality in elderly individuals who experience cognitive decline.5,6,48–50These studies have shown that diminishing cognitive performance over long intervals (7–9 yr)5,48or shorter intervals of time (≤ 2 yr)6,49,50is predictive of early death. It seems that the rate of cognitive decline is positively correlated with the risk of mortality, with rapid decline associated with the highest death rates.5,6The rate of cognitive decline in our postoperative patients could accurately be described as “precipitous” in comparison with that found in patients in these population-based studies. Moreover, there is evidence that a rapid decline in cognitive function has a significantly greater effect on mortality rates among the “young elderly” (<75 yr) than the “old elderly” (>75 yr).6,49,50Our elderly age group consisted predominantly of young elderly individuals (mean age, 70 yr) who may have been particularly vulnerable to the abrupt onset of cognitive dysfunction they experienced in the postoperative period.

It is unclear whether there is a causal link between POCD and mortality or if these outcomes are solely related to the patient’s underlying medical condition. In a previous publication based on the same study population, we reported that preoperative comorbidity was the major predictor of mortality after noncardiac surgery.51Patients in this current report who developed POCD were more likely to be older and to have a history of stroke—both conditions that are associated with greater comorbidity. Population-based studies are providing a growing body of evidence that cognitive decline may itself be a reliable indicator of the rate of aging and the risk of impending mortality.6Patients with prolonged POCD might be less adherent to medication or physical therapy regimens or may not recognize the need for medical follow-up if they experience symptoms of complications. These and other behaviors may result in suboptimal medical care and increase the patient’s risk of dying in the postoperative period. Patients’ perceptions of their own loss of intellectual function and reduced physical capacity may contribute to postoperative depression. Clinical depression is reported to be an independent and additive cofactor with cognitive impairment in determining the risk of mortality in the elderly.52 

Our previous publication focused on the evaluation of preoperative predictors and intraoperative predictors of postoperative outcomes.51The intention of this study was to find variables that could predict long-term outcome or whose modification could potentially improve outcome. A surprising finding in that publication was that cumulative deep hypnotic time increased the risk of death in the first year after surgery. POCD was not included in that analysis because it is a postoperative complication. We performed a post hoc  analysis on a composite bivariate endpoint, positive values of which indicated patients who either died within 1 yr of surgery or exhibited POCD at 3 months or both. In this analysis, the cumulative deep hypnotic time trended toward significance (P = 0.08). Future studies enrolling a larger number of patients are needed to determine whether cumulative deep hypnotic time is related to the occurrence of POCD.

A major limitation of all studies investigating POCD is that there are major differences in research methodologies, including the neuropsychological test batteries, the interval between testing sessions, the definition of POCD, and the statistical methods used to analyze the data.10These differences make it impossible to compare the magnitude of cognitive change in our patients with patients in other studies. However, our investigation used a robust study design consistent with recent guidelines and the methodology of the ISPOCD1 and ISPOCD2 studies, allowing us to compare our findings with the findings of these multinational studies.3,13The ISPOCD studies use age- and ability-matched control groups studied at the same intervals as patients to compensate for learning effects, and we followed these guidelines in our study.3,13The incidence of late POCD in the control subjects in our study was 1.8% in the elderly, 4.8% in the middle-aged, and 6.3% in the young. In contrast to expectation, the prevalence of POCD was lower in elders in our control group. However, the prevalence rates were not significantly different among the age groups and are in agreement with the POCD rates seen in controls in the ISPOCD studies (2.8% in the elderly and 4.1% in the middle-aged studies).3,13Clearly, these differences warrant further exploration.

The early testing point was to be completed at hospital discharge or 7 days after surgery, but some of the patients were too ill to be tested at 1 week, so they were tested as late as 1 month after surgery. The late testing point in this study was designed to occur at 3 months after surgery. Because many of the patients in our study lived long distances from the hospital, the late testing window was extended to 6 months after surgery, resulting in large SDs in the testing time (table 5). It is possible that additional cognitive improvement occurred in patients who were tested later than 1 week or 3 months after surgery, resulting in an underestimation of the prevalence of POCD. However, the median times of testing in our study were similar to the testing times for the elderly patients in the ISPOCD13and the middle-aged patients in the ISPOCD2 studies,13making the results of these studies comparable.

In conclusion, we found that increasing age, lower educational level, a history of previous cerebral vascular accident without residual impairment, and POCD at hospital discharge were independent risk factors for cognitive decline at 3 months after major noncardiac surgery. Although large numbers of patients of all ages were discharged from the hospital with POCD, only the elderly went on to have late cognitive decline. Of considerable concern, patients who experienced POCD at hospital discharge and at 3 months after surgery were more likely to die in the first year after surgery.

The authors thank Margaret Rosenberg, R.N., and Louise Paulus, C.R.N.A., for their assistance with data collection and coordination of the research protocol; Jan van der Aa, Ph.D., for data management and statistical support; and Vijaya Komaragiri, Information Technology Senior, for statistical support. These individuals are all on the faculty of the University of Florida, Gainesville, Florida. The authors also thank the members of the International Study of Postoperative Cognitive Dysfunction (ISPOCD) for their assistance with the design of this research protocol and the surgical, anesthesia, and nursing staff at Shands Hospital at the University of Florida, Gainesville, Florida, for their cooperation with this study.