Delayed psychomotor recovery or cognitive dysfunction after major surgery and anaesthesia are well documented. Risk factors for delayed psychomotor recovery or postoperative cognitive dysfunction (POCD) in the early postoperative period include advanced age , the magnitude of surgery, lower activity scores , duration of bed rest and hospitalization , sleep deprivation, postoperative pain, pre-existing disease  and the effects of sedative, analgesic or anaesthetic drugs [4,5]. However, only advanced age has been correlated with late (longer than 3-month duration) POCD .
POCD involves several components and its definition is not yet precise . Many psychometric tests have been used to measure different aspects of postoperative psychomotor recovery or POCD but most tests need specialized equipment or expertise for their performance or interpretation and are prolonged. Computerized cognitive function tests may be more sensitive than non-computerized tests but few are easily performed at the bedside [4,5]. Postoperative deficit in sustained attention is different from other aspects of POCD and little has been documented about it. A new computer-based test, the sustained attention to response task (SART) test, can detect deficits in sustained attention in patients with traumatic brain injury and correlates with self-reported slips in attention in normal volunteers [6,7] that are not detected by other tests of psychomotor function. Subjects do not need to be computer literate to perform the test [8,9]. In the absence of gross psychomotor dysfunction patients themselves may be unaware of a subtle impairment of sustained attention, and thus daily activities that need sustained attention such as cooking, operating machinery or driving may be potentially hazardous. We previously found that the SART test detected deficits in sustained attention after minor surgery at the time of discharge from hospital  but the significance of this after major surgery is not known. In this study we used the SART test to measure the magnitude and duration of deficits in sustained attention after major surgery. The SART test was also administered to an age and gender-matched control group, to establish whether repeated SART test performance would alter subsequent performance by a learning or practice effect.
Materials and methods
Following local research Ethics Committee approval and informed consent, 21 patients (ASA I-II) undergoing total abdominal hysterectomy under general anaesthesia were recruited to the study. Exclusion criteria were patient refusal, inability to perform the test due to problems with manual dexterity or visual acuity, pre-existing central nervous system disease or sedative drug medication. The investigator first gave a verbal explanation and a short demonstration of the test; all tests were conducted in a quiet, enclosed area. Patients completed the test preoperatively and at 24, 48 and 72 h postoperatively. No patients received premedication. Patients received a balanced anaesthetic technique comprising an induction dose of propofol and fentanyl 1-2 μg kg−1, maintained with isoflurane 1-2% in nitrous oxide (66%) and oxygen. Rocuronium or atracurium were used to produce neuromuscular blockade and intravenous (i.v.) morphine 0.1-0.2 mg kg−1 was administered during surgery. All patients received i.v. morphine via a patient-controlled analgesia system for 24 h postoperatively, after which analgesia comprised oral acetaminophen (paracetamol) 1 g 4-6 hourly and rectal diclofenac 100 mg daily. Patients' age, duration of surgery, untoward events during and after surgery, and time to discharge were recorded.
The SART test was also administered to a control group of 20 healthy age-matched female volunteers at similar time intervals (0, 24, 48 and 72 h) to assess for a possible learning effect during repeated performance of the test.
SART test and interpretation
The SART test comprises a series of 225 single numbers (1-9) displayed in a random order and size on a computer screen over 4.3 min. Each number is displayed for 250 ms and then masked by a ring with a diagonal cross in the middle for 900 ms. Participants are asked to respond by clicking on the number each time it appears, but to withhold their response if the number is ‘3’, with equal emphasis given to speed and accuracy of responses. The test included twenty-five ‘3’s distributed randomly throughout the 225 numbers. A short practice trial of 18 numbers is incorporated before each formal assessment. The number of errors made, that is, ‘3’s clicked on, and the number of non ‘3’s clicked on, is recorded. Averaged response times (RT) from appearance of a number on screen to clicking on the number are recorded for numbers (a) pre-correct response (i.e. response to ‘3’ correctly withheld), (b) pre-incorrect response (i.e. a ‘3’ incorrectly clicked on), and (c) post-incorrect response. The number of errors made is a sensitive indicator of a deficit in sustained attention. The number of non ‘3’s clicked on is a measure of the ability of subjects to attend to the test per se. A drift in sustained attention is demonstrated by comparing RT pre-correct response with RT pre-incorrect response, as subjects tend to respond progressively more quickly (i.e. a decreased RT) before a subsequent incorrect response. This decrease of RT before an incorrect response is predictive of the impending occurrence of a SART error. The normal response is that subjects slow down (i.e. RT increases) after an incorrect response when they realize an error has been made. If sustained attention is impaired as in severely head injured patients this pattern of increased RT after an incorrect response is not seen (i.e. there is no realization that an error has occurred) and the number of errors is also higher [8,9]. Defects in performance of the SART are correlated with deficits in activities of daily living in both normal volunteers and those with mild head injury [8,9].
Statistics and study power
Data were recorded and analysed using the SART software on a 486 MHz laptop computer. Statistical analysis was performed using general linear model analysis of variance for repeated measures with time and group as within and between subject factors, using SPSS for Windows computer software (release 11.01.2001). Results were considered significant if P < 0.05. Power calculation based on previous data  showed that a minimum of 19 subjects per group would be required to demonstrate a difference of three in the number of errors made (α = 0.05, β = 0.8).
Twenty-one patients were initially recruited to the study but one patient withdrew after 24 h so data from 20 patients and 20 age-matched females were analysed. Mean (SD) age was 48.6 (11.7) yr in the patient group and 48.4 (8.7) yr in the control group. Mean (SD) duration of surgery was 61.25 (11.4) min. No patients experienced any adverse events intra-operatively or postoperatively. Mean (SD) postoperative morphine consumption was 46.0 (19) mg over 24 h. Range of time to discharge was 4-7 days, with most patients discharged on the 5th post-operative day.
The results of the SART test are shown in Table 1 and Figures 1 and 2.
The number of errors showed a significant interaction between the groups over time (P = 0.023). Patients' error rates increased at 24 h postoperatively before reducing to preoperative values at 72 h, although these within-group differences were not statistically significant. In contrast, in the control group repeated SART test performance was associated with a learning effect as the number of errors decreased significantly at 48 and 72 h compared with initial values (P < 0.001) although between group differences were not statistically significant (P = 0.074 at 72 h). The number of correct responses (non ‘3’s clicked on) also demonstrated a significant interaction between the groups over time (P = 0.006). The number of correct responses decreased after surgery (P < 0.05 within group over time) before scores subsequently improved, whereas control subjects scored more correct responses with repeated performance on the test (P < 0.05 over time within group). The number of correct responses was significantly lower in the patient group at 24-48 h postoperatively (P < 0.05 between groups).
The overall RT (pre-correct response, pre- and post-error) decreased slightly over time in the control group (P = n.s.). However, all three RT indices slowed significantly at 24 h postoperatively in the patient group before subsequently decreasing, and the differences in RT pre-correct response, pre- and post-error between groups were statistically significant at 24-48 h (all P < 0.05). Patients' RT recovered with no differences between groups in RT indices at 72 h postoperatively. Both groups followed a typical pattern of a significant decrease in RT of 40-60 ms (i.e. speeding up) before making an error followed by a significant increase in RT by 70-100 ms (i.e. slowing down) after an error (Table 1). RT before an error was significantly quicker than RT before a correct response at all time points in both groups (P < 0.05) except at 48 h in the patient group (P = 0.054). Both groups behaved in a similar pattern and there were no significant interactions in the data between the groups.
In this study we found the number of errors made during performance of the SART test increased at 24 h after major gynaecological surgery under general anaesthesia, before decreasing to preoperative values at 72 h postoperatively. RT (before or after an error or before a correct response) also showed a similar pattern. Numbers of correct responses decreased at 24-48 h postoperatively. The return towards preoperative values after 48-72 h indicates that patients could attend to the test per se. In contrast, the number of errors and reaction times decreased significantly over 72 h in the control group, with significant interactions between the groups in number of errors made (P = 0.023) and number of correct responses (P = 0.006). This improvement in the control group's performance, that is a decreasing number of errors and overall quickened RTs is consistent with a learning effect. The different pattern in the patient group suggests impaired psychomotor performance after 24 h or impaired learning during the first 24-72 h after major surgery. Though RTs in the control group were quicker, there was no significant interaction between the groups. This suggests that by 48-72 h patients had sufficient gross psychomotor skills to attend to the test in the same way that the controls did. The number of correct responses also increased in the control group, although the absolute differences were small. This is partly explained by the duration of the SART test, so that using 225 digits of which 25 were a number ‘3’, the maximum possible number of correct responses was 200.
Sustained attention is defined as ‘the ability to self-sustain mindful, conscious processing of stimuli whose repetitive, non-arousing qualities would otherwise lead to habituation and distraction to other stimuli’. The SART test reliably detects sustained attention deficit, which correlates with both severity of head injury and also with self-reported slips in attention in normal volunteers . In this study, both groups showed the characteristic pattern of a decrease in RT (i.e. speeding up) before an incorrect response with an increase in RT (i.e. slowing down) after the incorrect response. This decrease in RT is predictive of a subsequent error, with the subsequent increase in RT occurring when the subject realizes that an error has been made. Subjects with a severe traumatic brain injury are unable to recognize their mistakes so RT before and after an error are similar [8,9]. The fact that RT slowed down after an error in both patients and controls suggests that they understood and performed the SART test correctly, and that these aspects of attentional processing are not affected by major surgery under general anaesthesia.
The results of this study are consistent with our previous data in patients undergoing day-case surgery under local or general anaesthesia, where we found an increased number of errors 2 h postoperatively in patients who were judged fit for discharge and had no obvious psychomotor deficits . RTs in the previous study were unaffected by minor surgery, but were slower in this study at 24 h after major surgery. Several factors could be responsible for this including residual effects of anaesthesia, morphine or its metabolites . However, the use of regional anaesthesia does not appear to decrease the incidence of POCD in elderly patients when compared with general anaesthesia  suggesting that non-pharmacological factors (including altered sleep cycles, environmental or emotional stress, bed rest, pain and anxiety) are important. Many patients complain of perceived cognitive impairment for a variable time following surgery, but subjective reports probably overestimate the incidence of POCD . Recent work investigating POCD in 40-60 yr olds demonstrated that 7 days postoperatively 19% of patients compared with 4% of controls showed POCD, but there were no significant differences after 3 months .
We recognize the study has limitations. Patient numbers were relatively low and patients were only followed up for 72 h, but by day 4 some patients were discharged from hospital. It would be valuable to continue future studies for a longer period, to determine the extent of improvement in test performance over several days or weeks, following the elimination of residual anaesthetic and analgesic drugs. Anaesthesia was standardized and therefore we cannot separate pharmacological effects from other factors (e.g. anxiety, stress, sleep disturbance). However, the aim of this study was to determine whether differences could be detected using the SART test after major surgery, and whether repeating the SART test on successive days would alter performance, both of which hypotheses have been confirmed. The effects of different types of surgery, anaesthetic techniques or drugs on sustained attention and the groups of patients at greatest risk are poorly defined. Future studies should also compare the SART test with self-reported and observer-reported attention failures in conjunction with other psychological tests used after major surgery. The exact reasons for the observed differences between the patient and control groups are not clear, but our findings suggest impaired psychomotor function and learning are involved.
In summary, the SART test was well accepted by patients, and could be administered by operators without formal psychological training. These data show that at 24-72 h after major surgery patients have a deficit in sustained attention compared with a control group, with a pattern of deficit less marked and different from that after severe head injury. The study also demonstrated that a significant learning or practice effect on repeating the test occurred in a control group but not in patients undergoing major gynaecological surgery. Further work should clarify the factors contributing to this impaired learning.
We wish to thank Professor Ian Robertson, Department of Psychology, Trinity College, Dublin, Ireland, who supplied the SART test and provided valuable advice on the project.
1. Moller JT, Cluitmans P, Rasmussen LS et al
. Long-term postoperative cognitive dysfunction in the elderly: ISPOCD1 study. Lancet
2. Johnstone T, Monk T, Rasmussen LS et al
. Postoperative cognitive dysfunction in middle-aged patients. Anesthesiology
3. Canet J, Raeder J, Rasmussen LS et al
. Cognitive dysfunction after minor surgery in the elderly. Acta Anaesthesiol Scand
4. Korttila KT. Post-anaesthetic psychomotor and cognitive function. Eur J Anaesthesiol
5. Bailie R, Christmas L, Price N, Restall J, Simpson P, Wesnes K. Effects of temazepam premedication on cognitive recovery following alfentanil-propofol anaesthesia. Br J Anaesth
6. Levin HS, High WM, Williams DH et al
. Dichotic listening and manual performance in relation to magnetic resonance imaging after closed head injury. J Neurol Neurosurg Psychiatr
7. Stuss D, Gow AD. Frontal dysfunction after traumatic brain injury. Neuropsychiatr Neuropsychol Behav Neurol
8. Robertson IH, Manly T, Andrade J, Baddeley BT, Yiend J. ‘Oops!’: performance correlates of everyday attentional failures in traumatic brain injured and normal subjects. Neuropsychologia
9. Manly T, Robertson IH, Galloway M, Hawkins K. The absent mind: further investigations of sustained attention to response. Neuropsychologia
10. Thompson JP, Rushman SC, Fox AJ, Lloyd AJ, Atcheson RA. Using the sustained attention to response test to detect cognitive dysfunction after day case surgery. Eur J Anaesthesiol
11. Meineke I, Freudenthaler S, Hofmann U et al
. Pharmacokinetic modelling of morphine, morphine-3 glucuronide and morphine-6-glucuronide in plasma and cerebrospinal fluid of neurosurgical patients after short-term infusion of morphine. Br J Clin Pharmacol
12. Begg A, Drummond G, Tiplady B. Assessment of postsurgical recovery after discharge using a pen computer diary. Anaesthesia
*These data were presented in part at the Anaesthetic Research Society meeting, Glasgow, April 2003.