Assessment of Vigilance Recovery in Rats Following Anesthesia Using a Touchscreen-Based Continuous Performance Task

Abstract

Background:

Cognitive recovery following general anesthesia is clinically important but incompletely understood. Rodent studies in anesthesia research have primarily examined memory and learning, while vigilance and other neurocognitive domains remain underexplored. To address this gap, we employed the rodent Continuous Performance Task (rCPT), a touchscreen-based translational paradigm, to characterize anesthetic-specific recovery trajectories of vigilance following anesthetic emergence.

Methods:

Sixteen adult Sprague Dawley rats (8 male, 8 female) were trained to proficiency on the rCPT. Once trained, loss of righting was induced by sevoflurane (3% for 20 min), propofol (10 mg/kg i.v. bolus), ketamine (50 mg/kg i.v. over 10 min), or dexmedetomidine (20 μg/kg i.v. bolus), delivered in a randomized order every other week. Animals were placed supine in the cognitive testing chamber and their performance was recorded for four hours. Return of righting reflex (RORR) was used to indicate anesthetic emergence. Vigilance recovery was defined as 20 consecutive trials with an average hit rate >0.6.

Results:

The latency to recover vigilance was delayed following ketamine and dexmedetomidine compared to baseline (Hazard Ratio in ketamine, 5.34; 95% CI, 2.07, 13.76; Hazard Ratio in dexmedetomidine, 5.33; 95% CI, 2.08, 13.62; both p<.0001). Reward retrieval latency recovered 4 hours after RORR following ketamine (percent change from baseline, −24.14%; 95% CI, −54.69, 17.89%) but not recover at the same time point following dexmedetomidine (percent change from baseline, −47.85%; 95% CI, −71.94, −20.23%).

Conclusions:

The rCPT identified delayed recovery of vigilance following ketamine and dexmedetomidine. Dexmedetomidine showed prolonged impairment across behavioral measures, including visual processing speed (hit latency) and reward retrieval, suggesting sustained effects on psychomotor function, motor coordination, and behavioral engagement that persisted for at least 4 hours after RORR. These findings support the utility of the rCPT as a translational paradigm for characterizing agent-specific and domain-specific trajectories of cognitive recovery following anesthesia.

9. Summary Statement:

The rodent Continuous Performance Test (rCPT) revealed anesthetic-specific delays in vigilance recovery, with ketamine and dexmedetomidine causing prolonged sedative effect, while sevoflurane and propofol allowed rapid restoration of vigilance relatively.

Introduction

Understanding how cognition recovers following general anesthesia carries both scientific significance and clinical relevance for optimizing patient care in the immediate postoperative period. While rodent models have been invaluable for assessing cognitive function after various anesthetic regimens, most studies have relied on traditional maze-based tasks of learning and memory, such as the Morris Water Maze¹ and the Radial Arm Maze,² which have limited translational relevance. In contrast to the high-stakes survival contexts inherent in many rodent tasks, human cognitive assessments typically rely on simple report-based measures or performance in domain-specific perceptual tasks.^{3, 4} The advent of more sophisticated touchscreen-based operant paradigms for rodents now provides an opportunity to investigate cognition under conditions that more closely parallel clinical testing environments.

Operant conditioning tasks, widely used in cognitive research, have only recently been applied in anesthesia rodent studies. These paradigms have been used to assess sustained attention, working memory, and spatial memory in both the early and later stages of anesthetic recovery in rodents.^5–7 In parallel, research in healthy human volunteers has recently shown that recovery of cognitive function following general anesthesia is gradual and domain-specific. Notably, Mashour et al. reported that vigilance – defined as alertness to one’s surroundings⁸ – was among the slowest cognitive domains to recover following emergence from isoflurane anesthesia, highlighting the need to further characterize vigilance recovery across different anesthetic agents.⁹ Although recovery of vigilance is clinically important for ensuring the safe discharge of patients following surgery, its time course has not been systematically characterized across different classes of anesthetic and sedative agents. Therefore, the application of analogous behavioral paradigms in animal models provides a valuable translational approach for characterizing agent-specific trajectories of vigilance recovery and identifying factors that influence post-anesthetic cognitive recovery.

In this study, we employed the rodent Continuous Performance Task (rCPT) to characterize the time course of vigilance recovery following exposure to various anesthetic agents. The human CPT, originally designed to detect attentional impairments in disorders such as Alzheimer’s disease^{10, 11} and schizophrenia,¹² requires subjects to discriminate between sequentially presented visual target and non-target stimuli. The rCPT closely replicates the structure and attentional demands of the human version and has been validated for use in animal models and pharmacological studies.^13–15 The rCPT utilizes non-target stimuli that share overlapping perceptual features (e.g., brightness, contrast, color) with target stimulus, more closely resembling human vigilance paradigms in which sensitivity to subtle perceptual differences is critical.¹⁶ These characteristics led us to hypothesize that the rCPT may provide good sensitivity in detecting vigilance impairments following anesthetic exposure.

Using this approach, we investigated vigilance performance in rats following exposure to mechanistically distinct anesthetic and sedative agents. The goal of this study was to characterize agent-specific trajectories of vigilance recovery following RORR using the rCPT. Through this framework, we aimed to describe how vigilance recovers over time in each condition and to evaluate the rCPT as a translational paradigm for studying post-anesthetic cognitive recovery.

Methods

All experimental procedures were approved by the Institutional Animal Care and Use Committee of Massachusetts General Hospital and were conducted in accordance with the Guide for the Care and Use of Laboratory Animals¹⁷. This study adheres to the ARRIVE guidelines.¹⁸

Animals

Male and female Sprague–Dawley rats obtained from Charles River Laboratories (Wilmington, MA) underwent training at 7–8 months of age (280–540 g). Animals were housed in pairs under a 12-hour light/dark cycle (lights on at 7:00 AM) in a temperature-controlled environment, with ad libitum access to water. After arrival, rats were acclimated for at least one week, during which they were handled daily. Once training commenced, animals were food-restricted five days per week and received 4 g (females) or 5 g (males) of food pellets after each daily training session. Daily weight monitoring ensured body weight remained above 85% of free-feeding levels. No animals were excluded from the final analysis.

Behavioral Task

Behavioral training was conducted in Bussey-Saksida Rat Touch Screen Chambers (Lafayette Instrument, IN, USA), controlled via the ABET II interface and software. Rats were trained on the Rodent Continuous Performance Task (rCPT), a modified task based on human continuous performance tasks, following the developer’s instructions.¹³ Training and testing occurred in a trapezoidal room equipped with a touch screen on one side and a magazine food tray on the opposite side (Fig. 1A).

Fig. 1:

Overview of the experimental procedure and Rodent Continuous Performance Task (rCPT). (A) Top view of operant chamber showing the stimulus presentation window and reward port. (B) The rodent engages in the task by approaching the stimulus window. (C) The rodent retrieves reward pellet from the reward port. (D) Representative image of target (S+) stimulus and non-target (S-) stimuli. (E) Four possible response types are shown: Responses to S+ are scored as either ‘hit’ or ‘miss’; responses to S- are recorded as either ‘false alarm’ or ‘correct rejection’. (F) Experimental timeline: Following task acquisition to criterion, rats completed two days of task maintenance, a baseline test, an anesthetic test, and a 1-day post-test. Both the baseline and post-tests were identical to maintenance sessions (30 minutes each). On the anesthetic test day, the drug was administered and animals were placed supine in the operant chamber. RORR was recorded and testing continued for up to 4 hours post-RORR or until 150 hit-trials were completed, whichever occurred first. This procedure was repeated biweekly, with the administered agent randomly assigned for each experimental week.

Each trial began with the presentation of a white square frame (3 × 3 cm, 3.5 cm above the floor), indicating the response zone. A 45 mg grain pellet (dustless Precision Pellets, Bio-Serv, F0165) was delivered to initiate the session, accompanied by tray light illumination. Once the pellet was retrieved, the tray light was extinguished and a 2–3 second inter-trial interval was initiated (ITI; randomized).

Following the ITI, either a target stimulus (S+, horizontal stripes, Fig. 1D) or one of four non-target stimuli (S−) was displayed within the response frame for 2 seconds. The stimulus presentation probability was 33.3% for S+ and 66.7% for S−. As illustrated in Fig. 1E, touches to S+ were scored as ‘Hits’ and were followed by reward delivery, light, and tone (3 kHz). After a 2 second pause to allow the rat to eat the reward pellet, the next ITI began. If no touch was made during S+ or S− presentation, a ‘Miss’ or ‘Correct Rejection’ was recorded, respectively. Touches to S− were scored as ‘False Alarms’ and triggered a correction trial, in which only S− stimuli were presented until the animal withheld responses, ending with a ‘Correction Trial Correct Rejection’.

Anesthetic Testing

Each rCPT training session lasted 30 minutes. Proficiency was defined as an average hit rate (= number of hits / (number of hits + number of misses)) greater than 0.7 across two consecutive sessions. Most animals reached this criterion within 14 sessions. After achieving proficiency, all animals underwent 2 additional training sessions, followed by a 30-minute baseline session (Fig. 1F). Anesthetic agent exposure and post-anesthetic behavioral tests were performed the day after the baseline session. Post-anesthetic behavioral testing lasted 4 hours after return of the righting reflex (RORR), or until 150 hits were reached. Rodents were also assessed on the rCPT one day following anesthetic exposure for 30 minutes to test for any persistent effects of the drug.

Drugs

Animals were administered the following anesthetic or sedative agents in a randomized order: sevoflurane (20 minutes at 3%; Sandoz Inc., Princeton, NJ), propofol (10mg/kg IV bolus; Fresenius Kabi USA LLC, Lake Zurich, IL), ketamine (50mg/kg infused over 10min; Hospira Inc., Lake Forest, IL), or dexmedetomidine (20μg/kg IV bolus; Eugia US LLC, East Windsor, NJ). Doses were chosen based on those previously shown to reliably induce loss of the righting reflex, a well-established proxy for loss of consciousness.⁶

Post-anesthetic testing

Each rat received one drug every 2 weeks for a total of 4 exposures per animal. Intravenous drugs were administered via a 24-gauge catheter placed in the lateral tail vein under brief isoflurane anesthesia (typically 2–5 minutes). Animals fully recovered from isoflurane for at least 45 minutes prior to administration of the intravenous drug. After administering the drug, the catheter was flushed with 0.5ml saline and removed. Animals were then placed supine in the cognitive testing chamber. RORR was determined from recorded video. All recordings were randomized and evaluated by a scorer blind to the experimental drug. Following RORR, each animal was tested until they reached 150 hits, or a maximum time limit of 4 hours.

Statistics

Statistical analyses were performed using GraphPad Prism 10.4 (Dotmatics, Boston, MA) and R (version 4.0.3). A total of 16 rats (8 males, 8 females) were included. Sample size estimation was based on a previously observed sex difference in time to RORR (Cohen’s d ~ 1.8).¹⁹ Power analysis using G*Power (v3.1.9.6; Faul et al., 2007) indicated that a sample size of seven per group would provide sufficient power to detect sex differences using a two-tailed Mann-Whitney U test at a type I error rate of 5%.²⁰ An additional animal per group was included to account for potential attrition or failure to reach task criteria, resulting in a final sample size of eight per group.

Primary outcomes included hit rate, false alarm rate, sensitivity index, responsivity index, time to RORR and time to vigilance recovery. The sensitivity index is a nonparametric measure of an individual’s ability to distinguish signal from background noise. ²¹ Values range from 0 to 1, where 0 indicates chance-level discrimination between target and non-target stimuli, and 1 indicates perfect discrimination, defined as correct responses to all target stimuli without responses to non-target stimuli. The responsivity index reflects response tendency, with lower values indicating a more conservative response bias and higher values indicating a more liberal response bias.

$$Hitrate=\frac{Numberofhits}{Numberofhits+Numberofmisses}$$ #(1)

$$Falsealarmrate=\frac{Numberoffalsealarms}{Numberoffalsealarms+Numberofcorrectrejections}$$ #(2)

$$Sensitivityindex=\frac{Hitrate‐Falsealarmrate}{2\mathrm{*}\left(Hitrate+Flsealarmrate\right)‐{\left(Hitrate+Falsealarmrate\right)}^2}$$ #(3)

$$Responsivityindex=\frac{Hitrate‐Falsealarmrate‐1}{1‐{\left(Falsealarmrate‐Hitrate\right)}^2}$$ #(4)

To assess potential carryover effects, linear mixed-effects models were additionally applied to evaluate the main effect of exposure order. Drug condition and exposure order, along with their interaction, were included as fixed effects, and subject was included as a random intercept. Analyses were performed in R using the lme4 package.

Vigilance recovery is defined as the interval between RORR and achieving a hit rate above 0.6. This threshold was determined based on initial task performance from proficient rats without anesthesia, whose average hit rate exceeded 0.6 in the first 20 trials. Vigilance recovery trajectories (cumulative incidence curves) were compared between baseline and intervention sessions using the log-rank (Mantel-Cox) test.

Secondary outcome variables included total number of trials to reach a hit rate >0.8, as well as median hit latency and reward retrieval latency. We calculated normalized hit latency and reward retrieval latency relative to baseline values using the following equation (5).

$$\begin{array}{c}Normalizedlatency\left(\mathrm{\%}\right)=\frac{Observedlatency}{AveragelatencyfromBaseline}\times 100\end{array}$$ #(5)

Normalized hit latency was defined as the percent change in correct response latency relative to baseline. This measure reflects response speed during the rCPT, requiring subjects to discriminate between target and non-target stimuli, and thus serves as an index of vigilance-related processing time within the visual domain.

Normalized reward retrieval latency reflects response timing following an auditory cue (the tone feedback associated with reward delivery). This measure additionally incorporates motor movements, as subjects must disengage from the touchscreen, traverse chamber, and reach to reward port, thereby capturing psychomotor vigilance and motor coordination.

We hypothesized that these complementary measures capture distinct aspects of vigilance: hit latency reflects response timing associated with vigilance-related processing, whereas reward retrieval latency reflects psychomotor response involving coordinated movement, thereby providing a translational framework for interpreting anesthetic effects on vigilance across visuomotor domains.

For ketamine and dexmedetomidine, these measures were computed at 30-minute intervals to depict recovery trajectories, resembling the approach used in the human study by Mashour et al.⁹ For sevoflurane and propofol, values were computed at 10-minute intervals, as subjects completed 150 hits within 80 minutes after drug administration.

All data were assessed for parametric assumptions and those which failed were analyzed using nonparametric tests, as indicated. The effect of sex on RORR latency and vigilance recovery latency for each drug was assessed using unpaired t-tests whereas vigilance recovery latencies were compared between sexes using the Mann–Whitney U test due to non-normal distribution. Baseline performance comparison between sexes over seven weeks was assessed with two-way repeated-measures ANOVA with Geisser–Greenhouse correction. Changes in hit rate, hit latency and reward retrieval latency across three time points (baseline day, intervention day, and 24 hours post-intervention day) were analyzed using one-way repeated-measures ANOVA with Geisser–Greenhouse correction. Detailed results of post hoc multiple-comparison analyses are provided in the Supplemental Digital Content (Supplementary Tables 1–4).

Normality was tested using the D’Agostino-Pearson test. Time course data for normalized hit latencies and normalized reward retrieval latencies were evaluated with one-sample Wilcoxon tests due to small and variable sample sizes across time points and deviations from normality. We measured Pearson’s correlation to examine associations between primary and secondary outcomes for each agent. Data are presented as median [1st quartile-3rd quartile].

Results

Baseline Performance

All animals achieved task proficiency and underwent all anesthetic testing conditions. The median number of training sessions required to achieve proficiency was 8 [7.5–9.25], with no significant difference between sexes (p = .829). Baseline performance, assessed the day prior to anesthetic exposure, did not differ between sexes with respect to hit rate, reward retrieval latency, nor latency to reach a hit rate >0.6 across the seven-week experimental period (Supplementary Fig. 1).

Sevoflurane

Sex did not affect time to RORR latency (p=.621; Fig. 2A) nor vigilance recovery latency (p=.505; Fig. 2B) following 20 minutes of 3% sevoflurane anesthesia. Median latency to RORR was 4.95 [2.96–6.66] minutes in females and 5.59 [1.65–6.08] minutes in males. Median time to achieve a hit rate >0.6 was 5.08 [3.66–7.67] minutes for females and 5.7 [4.47–15.02] minutes for males. Compared to baseline performance, sevoflurane exposure delayed vigilance recovery significantly (median time to recovery in baseline, 2.34 [1.93–4.61] minutes; post-sevoflurane, 5.03 [3.84–8.9] minutes; hazard ratio, 2.356; 95% CI of ratio, 1.10, 5.02; p=.008; Fig. 2C), with the slowest animal recovering 40 minutes following righting. Sevoflurane did not significantly affect hit rate but reduced false alarm rate compared to baseline (Fig. 2D-E). Sensitivity index, Responsivity index, hit latency, and reward retrieval latency were not affected after exposure compared to baseline (Fig. 2F-I; see Supplementary Table 1 for all post hoc multiple comparison results). RORR latency did not correlate with the number of trials required to reach a hit rate >0.8 (p = .822; r = −0.19; 95% CI, −0.54, 0.44; Fig. 2J). Time-course analysis showed no significant change in hit latency across 10-minute windows; however, reward retrieval latency remained elevated until 30 minutes after RORR (Fig. 2K-L). Animals completed 150 trials in 72.46 [64.29–82.69] minutes after RORR.

Fig. 2:

RORR, vigilance recovery, and various types of metrics from sevoflurane inhalation. (A) Time to RORR did not differ significantly between sexes. (B) No sex differences in time to vigilance recovery. (C) Time to reach average hit rate >0.6 was significantly prolonged after sevoflurane inhalation, with recovery occurring up to 40 minutes after RORR. (D) Hit rates were unchanged between time points. (E) False alarm rates were reduced after sevoflurane inhalation but returned to baseline the next day. (F) Sensitivity indexes were unchanged between time points. (G) Responsivity indexes were not significantly altered during the test following sevoflurane inhalation compared to baseline test but were increased 24 hours after sevoflurane compared with intervention day. (H) Hit latencies after sevoflurane inhalation did not differ between baseline test and test after sevoflurane inhalation (intervention day) but were reduced 24 hours post-sevoflurane compared with baseline. (I) Reward retrieval latencies were not significantly altered during the test following sevoflurane inhalation compared to baseline test but were reduced 24 hours after sevoflurane compared with intervention day. (J) RORR latency and number of trials to achieve >0.8 hit rate did not show strong correlation. (K) Normalized hit latencies after RORR were not significantly different from baseline level (dashed line; 0%). The final time point was set at 80 minutes post-RORR, as subjects required 63.3 [58.6–79.5] minutes to complete 150 trials following RORR after sevoflurane inhalation. (L) Normalized reward retrieval latencies were significantly prolonged until 30 minutes after RORR compared to baseline level. ‘*’ indicates p<.05, ‘**’ p<.01, ‘***’ p<.001, and ‘ns’ not significant, n=16.

Propofol

Sex did not affect time to RORR latency (p=.414; Fig. 3A) nor vigilance recovery latency (p=.798; Fig. 3B) following 10 mg/kg propofol administration. Median RORR latency was 8.45 [7.37–12.88] minutes for females and 11.99 [9.00–14.17] minutes for males. The latency to reach a hit rate >0.6 was 11.64 [5.22–18.88] minutes in females and 10.89 [5.87–15.50] minutes in males. Surprisingly, compared to baseline, propofol did not significantly delay latency to vigilance recovery (median time to recovery in baseline, 2.37 [1.93–4.8] minutes; post-propofol, 4.05 [2.33–8.05] minutes; hazard ratio, 1.598; 95% CI of ratio, 0.78, 3.25; p = .169; Fig. 3C). Hit rate, false alarm rate, sensitivity index, responsivity, and hit latency were also unaffected (Fig. 3D-H; see Supplementary Table 2 for all post hoc multiple comparison results). Reward retrieval latency was modestly increased following exposure compared with baseline. Although the absolute difference was small (mean difference, 76 ms; 95% CI, 19, 132 ms; Fig. 3I), the standardized within-subject effect size was large (Cohen’s dz = 0.88), indicating a consistent shift across subjects. RORR latency did not correlate with the number of trials required to reach a hit rate >0.8 (p = .479; r = −0.19; 95% CI, −0.64, 0.35; Fig. 3J). Time-course analysis showed no changes in hit latency throughout the session (Fig. 3K), while reward retrieval latencies remained elevated for 30 minutes after RORR (Fig. 3L). Animals completed 150 trials in 74.69 [69.59–92.25] minutes after RORR.

Fig. 3:

RORR, vigilance recovery, and various types of metrics from propofol injection. (A) No sex differences in RORR latencies. (B) Vigilance recovery latencies did not differ significantly between sexes (C) All animals reached hit rate >0.6 within 20 minutes after RORR. (D) Hit rates, (E) False alarm rates, (F) Sensitivity indexes, and (G) Responsivity indexes were not significantly changed after exposure to propofol compared to baseline level. (H) Hit latencies were not significantly different during the test after propofol injection when compared to baseline. (I) Reward retrieval latencies were elevated post-propofol compared to baseline. (J) RORR latency and number of trials to achieve >0.8 hit rate did not show strong correlation. (K) Normalized hit latencies after RORR were unchanged from baseline level (black dashed line; 0%). The final time point was set at 80 minutes post-RORR, as subjects required 69.3 [58.0–84.5] minutes to complete 150 trials following RORR after propofol injection. (L) Normalized reward retrieval latencies were significantly prolonged for up to 30 minutes post-RORR compared to baseline level. ‘*’ indicates p<.05, ‘**’ p<.01, ‘***’ p<.001, and ‘ns’ not significant, n=16.

Ketamine

Sex did not affect RORR latency (p=.422; Fig. 4A). However, vigilance recovery latency was significantly longer in females compared with males following 50 mg/kg ketamine infusion (mean difference, 47.69 ± 13.46 minutes; 95% CI, 18.36, 77.01; p=.017; Fig. 4B). One female subject did not reach the 0.6 hit rate within 4 hours after RORR, and one male subject was excluded as an outlier (latency = 213 minutes), exceeding the upper bound based on the 1.5xIQR rule. Median RORR latency was 24.72 [19.83–37.33] minutes for females and 23.47 [17.82–31.38] minutes for males. The latency to reach a hit rate >0.6 was 131.95 [102.78–152.06] minutes in females and 63.36 [56.62–99.5] minutes in males. Overall, exposure to ketamine significantly delayed vigilance recovery (median time to recovery in baseline, 2.12 [1.91–3.21] minutes; post-ketamine, 102 [77.1–143.1] minutes; hazard ratio, 5.347; 95% CI of ratio, 2.07, 13.76; p < .0001; Fig. 4B), reduced hit rate, false alarm rate, sensitivity index, and responsivity index (Fig. 4D-G; see Supplementary Table 3 for all post hoc multiple comparison results). Ketamine also increased hit latency (Fig. 4H) and reward retrieval latency (Fig. 4I). One subject with a reward retrieval latency of 17 seconds in the intervention session was identified as an outlier and excluded based on the 1.5×IQR criterion. RORR latency did not correlate with the number of trials required to reach a hit rate >0.8 (p = .507; r = 0.202; 95% CI, −0.39, 0.67; Fig. 4J). Time-course analysis revealed that hit latency remained elevated for up to one hour post-RORR (Fig. 4K). Intriguingly, reward retrieval latency remained impaired throughout the session (Fig. 4L). Notably, none of the animals collected rewards within the first 30 minutes post-RORR, and three subjects did not retrieve their rewards between 0.5 and 1 hour post-RORR. This suggests that animals were able to detect and respond to stimuli but failed to recognize or access the reward promptly. Seven animals completed 150 trials within the 4-hour post-RORR period (median number of hits = 136 [123.25–147]).

Fig. 4:

RORR, vigilance recovery, and various types of metrics following ketamine infusion. (A) Time to RORR did not differ significantly between sexes. (B) Female rats exhibited longer vigilance recovery latency compared to males. (C) Time to achieve >0.6 hit rate was significantly prolonged after ketamine infusion, and reaching up to 3.5 hours after RORR. (D) Hit rates were significantly decreased after ketamine infusion and recovered the following day. (E) Ketamine significantly reduced false alarm rates during testing which is not normalized by the following day. (F) Sensitivity indexes and (G) Responsivity indexes were significantly decreased after ketamine infusion and recovered the following day. (H) Hit latencies were elevated after ketamine infusion, but recovered to baseline level by the next day. (I) reward retrieval latencies were significantly prolonged after ketamine infusion but recovered by the following day. (J) RORR latency and number of trials to achieve >0.8 hit rate did not show strong correlation. (K) Time-course analysis showed delayed hit latencies up to 1 hour post-RORR after ketamine infusion compared to baseline. (L) Reward retrieval latencies were significantly prolonged compared to baseline. ‘*’ indicates p<.05, ‘**’ p<.01, ‘***’ p<.001, ‘****’, p<.0001, and ‘ns’ not significant; n=7 females and 7 males in (B); n=15 in (I); n=16 for all other panels.

Dexmedetomidine

Sex did not significantly affect RORR latency (p = .714; Fig. 5A) nor vigilance recovery latency (p = .710; Fig. 5B) following 20 μg/kg dexmedetomidine administration. Median RORR latency was 70.23 [59.42–111.74] minutes for females and 78.07 [67.10–95.91] minutes for males, which was not significantly different. Median vigilance recovery latency was 125.04 [107.29–140.16] minutes in females and 130.56 [83.80–149.99] minutes in males. Latency to attain >0.6 hit rate was significantly longer after dexmedetomidine compared to baseline (median time to recovery in baseline: 2.05 [1.9–2.87] minutes; post-dexmedetomidine: 135.1 [93.9–149.9] minutes; hazard ratio: 5.328; 95% CI of ratio, 2.08, 13.62; p<.0001; Fig. 5B), and two animals did not reach criterion within 4 hours after RORR. Dexmedetomidine significantly reduced overall hit rate and false alarm rate (Fig. 5D-E; see Supplementary Table 4 for all post hoc multiple comparison results). Sensitivity index was not affected but responsivity index was reduced after exposure to dexmedetomidine compared to baseline (Fig. 5F-G). Dexmedetomidine increased both median hit latency and reward retrieval latency (both p < .0001; Fig. 5H and 5I). RORR latency negatively correlated with the number of trials required to reach a hit rate >0.8 (p = .022; r = −0.65; 95% CI, −0.89, −0.12; Fig. 5J). Time-course analysis showed substantial variability in both hit and reward retrieval latencies during the first 2.5 hours post-RORR, followed by consistent impairment in the final stages (Fig. 5K–L). Six animals completed 150 hits within 4 hours after RORR (median number of hits within four hours after RORR was 120 [98.25–150]).

Fig. 5:

RORR, vigilance recovery, and various types of metrics after dexmedetomidine injection. (A) Time to RORR did not differ by sex. (B) No sex differences in vigilance recovery latencies. (C) 14 out 16 rats reached >0.6 hit rate, and the latency was significantly longer compared to baseline. (D) Hit rates were significantly reduced during testing which is normalized by the following day. (E) False alarm rates were significantly reduced in dexmedetomidine condition. (F) Sensitivity indexes were lower following dexmedetomidine administration, but the difference was not statistically significant. (G) Responsivity indexes were significantly reduced during testing which is normalized by the following day. (H) Hit latencies, and (I) reward retrieval latencies were significantly impaired post-dexmedetomidine compared to baseline, but returned to baseline the next day. (J) Time to RORR negatively correlated with the number of trials required to reach >0.8 hit rate. (K) Hit latencies remained highly variable for up to 2.5 hours following RORR and were subsequently prolonged to a statistically significant degree relative to baseline. (L) Reward retrieval latencies remained below baseline for the entire test session. ‘*’ indicates p<.05, ‘**’ p<.01, ‘***’ p<.001, ‘****’, p<.0001, and ‘ns’ not significant, n=16.

Discussion

In this study, we characterized vigilance recovery trajectories in rats following four distinct drug-induced unconscious states using the rodent Continuous Performance Task (rCPT). Vigilance rapidly returned after exposures to propofol and sevoflurane, whereas recovery was markedly delayed following dexmedetomidine and ketamine. Interestingly, RORR latency after dexmedetomidine negatively correlated with the number of trials required to reach a hit rate >0.8. Hit (correct response) latency remained unaffected by propofol and sevoflurane, although reward retrieval latency was transiently prolonged for up to 30 minutes post-RORR. In contrast, ketamine and dexmedetomidine significantly impaired hit rate and increased both hit latency and reward retrieval latency. Ketamine and dexmedetomidine both impaired responsivity index compared to baseline, but sensitivity index is significantly reduced in ketamine only. Consistent with findings in other studies, no long-term deficits were observed with any drug, and all animals returned to baseline performance within 24 hours.^5–7 There was no effect of drug exposure order across the experiment (see Supplementary Table 5 for detailed results from the linear mixed-effects model analysis).

Interestingly, each anesthetic agent produced a distinct cognitive recovery profile. Vigilance returned within 20–40 minutes post-RORR following sevoflurane or propofol exposure, whereas vigilance recovery was markedly delayed after ketamine and dexmedetomidine, extending up to several hours. These differences are consistent with the pharmacokinetics of elimination for each drug. Vincent et al. (2021) used the five‑choice serial reaction time task (5‑CSRTT), which assesses attentional control, working memory, and response inhibition, to characterize recovery trajectories after exposure to isoflurane, sevoflurane, propofol, ketamine, and dexmedetomidine.⁶ Their study demonstrated that return of the righting reflex (RORR) did not predict recovery of sustained attention, underscoring a dissociation between arousal and higher‑order cognitive function. Similarly, Obert et al. (2025) employed the paired-associates learning (PAL) task to probe long‑term visuospatial associative memory and found anesthetic‑specific delays in cognitive recovery, including a positive correlation between time to RORR and time to cognitive recovery after ketamine.⁷ Human studies provide a valuable reference, as Mashour and colleagues reported prolonged vigilance impairment up to 3 hours after recovery of consciousness (ROC) in healthy volunteers who received 3 hours of 1.3x the age-adjusted minimum alveolar concentration (MAC) of isoflurane.⁹ Participants in that study completed six cognitive test batteries, including psychomotor vigilance task (PVT), at 30-minute intervals to map recovery trajectories. Sprague-Dawley rats in our study regained vigilance more rapidly than human participants, but this likely reflects the shorter duration of sevoflurane exposure (20 minutes at 3%) and its faster pharmacokinetics relative to isoflurane. Although direct comparison between our rodent results and the human data is not feasible, the rCPT’s ability to detect post-anesthetic delays in vigilance recovery suggests that the touchscreen approach represents a valuable translational tool for investigating vigilance recovery.

One intriguing finding was the immediate recovery of vigilance following propofol anesthesia. Although propofol, at the dose used, produced a longer duration of unconsciousness compared to sevoflurane (mean difference = 6.17 minutes; 95% CI 3.8, 8.4; p<0.0001), animals showed no delay in vigilance recovery following propofol compared to baseline. This suggests that animals emerge from propofol in a more alert state, and while gross movement (e.g., reward retrieval) may be slowed, attentiveness and reaction times are intact. This feature highlights the utility of propofol for ambulatory procedures where quick cognitive recovery following emergence is particularly desirable.²² Recapitulating this feature in an animal model speaks to both the utility of this assay in assessing clinically relevant endpoints but also opens the door to probing the mechanisms underlying rapid recovery.

In contrast to propofol and sevoflurane, we observed sex difference in vigilance recovery following ketamine exposure. This finding is consistent with prior preclinical work, in which female rats exhibited a markedly prolonged latency to cognitive recovery despite similar RORR latency compared with male rats.⁷ Ketamine also produced marked and prolonged impairments in reward retrieval latency compared to baseline. Notably, shortly after emergence, rats frequently fail to retrieve the reward despite a correct response. As shown in Fig. 4K and Fig. 4L, two subjects made correct responses at 0.5 hours without subsequent reward retrieval, and only 2 of 12 subjects retrieved the rewards at the 1-hour time point. Although hit latency normalized within 90 minutes, animals remained slow to collect rewards for the entire session and maintained a low hit rate. The preserved reaction time (hit latency) combined with delayed or omitted reward retrieval may reflect a ketamine-induced dissociative state characterized by reduced responsiveness to external stimuli. Dávila et al. demonstrated that with increasing doses of ketamine associated with progressive stabilization of cortical dynamics, marked by a reduced fluctuations in the higher-frequency (above 20 Hz) EEG activity following neuronal perturbations, consistent with a dissociative state.²³ This stabilization has been proposed as a mechanism underlying altered perceptual processing,^{24, 25} including the emergence of hallucinatory states.²⁶ In line with this, the uncollected reward pellets within 1 hour post-RORR and the prolonged reward retrieval latency observed in this study may reflect a hallucinatory state or impaired perception process.

Dexmedetomidine administration produced significant and protracted impairments in vigilance and delayed reaction time. Although reward retrieval latencies recovered by the following day, they remained elevated throughout the test session. Likewise, hit latencies were persistently slowed, indicating that animals remained in a sedated state long after RORR. Comparable effects have been described in rodent studies employing the 5-choice serial reaction time task (5-CSRTT), a well-established paradigm of sustained attention. Sirviö et al. (1994), for instance, demonstrated that dexmedetomidine (0.3–3 μg/kg, subcutaneously) markedly increased omission errors and reduced premature responses without affecting accuracy, suggesting a reduction in overall arousal and task engagement rather than in perceptual discrimination.²⁸ In contrast, the α2-adrenoceptor antagonist atipamezole exerted a mild stimulant-like effect, increasing premature responses while sparing accuracy.²⁸ These convergent findings support the view that norepinephrine originating from the locus coeruleus is critical for sustaining attention and arousal.²⁹ Consistent with these preclinical results, a human study using dexmedetomidine (a 10-minute loading infusion followed by a 4-hour maintenance infusion) found that the drug slowed response speed without affecting accuracy in vigilance, attention, or memory tasks.³⁰ In line with this observation, the sensitivity index in our study was not significantly altered following dexmedetomidine exposure, despite a reduction in hit rate. This suggests that the lower hit rate may reflect residual sedative effects influencing response behavior rather than impaired stimulus discrimination. Collectively, the present data reinforce that the prolonged attentional deficits observed after dexmedetomidine reflect a state of decreased arousal rather than impaired sensory discrimination, highlighting the importance of noradrenergic modulation in sustaining vigilance.

Interestingly, a significant negative correlation between RORR latency and the number of trials required to reach a hit rate greater than 0.8 was observed only following dexmedetomidine administration. In other words, animals with shorter RORR latency required more trials to achieve high-performance criteria, whereas those with longer RORR latency reached this level more rapidly. One possible explanation is that animals with longer RORR latency regained the righting reflex at a later point along the dexmedetomidine washout trajectory and are therefore in a more advanced state of recovery when behavioral testing begins. Accordingly, this relationship may reflect residual sedative effects influencing arousal and behavioral engagement, rather than a distinct cognitive mechanism.

We observed prolonged increases in both normalized hit latency and normalized reward retrieval latency following dexmedetomidine exposure, persisting for at least four hours relative to baseline. Normalized hit latency reflects response speed during the rCPT and is influenced by visual processing as well as fine motor execution, whereas reward retrieval latency represents a composite behavioral measure incorporating psychomotor function, motor coordination, and motivational engagement. The concurrent delays observed across these complementary measures suggest that dexmedetomidine produces a sustained impairment affecting multiple components of task performance, including sensory processing speed, motor function, and goal-directed behavior, rather than reflecting a deficit in a single cognitive domain. Collectively, these findings are consistent with prolonged sedative effects that broadly impact arousal and behavioral responsiveness at the administered dose.

Several limitations should be noted. First, although return of righting reflex (RORR) was used as a standardized behavioral marker of emergence to assess recovery relative to a common functional threshold, differences in pharmacokinetic and pharmacodynamic profiles across anesthetic agents resulted in variability in the duration and persistence of drug effects. As such, animals may have been in different neurobiological states along their recovery trajectories despite being measured relative to RORR, which limits the interpretability of direct cross-agent comparisons. Therefore, the present findings should be interpreted as reflecting agent-specific patterns of vigilance recovery at the administered doses rather than equivalent stages of recovery across agents. Second, our operant conditioning paradigm depends on food-motivated behavior. Although drug-induced changes in appetite could theoretically affect performance, we previously confirmed that motivation for food reward remains intact following drug exposure.⁶ Lastly, unlike human studies in which multiple cognitive domains can be assessed within a single session, rodents can be trained on only one touchscreen task per animal. This single‑task constraint substantially increases the number of animals, time, and costs required to evaluate multiple cognitive domains, limiting experimental efficiency and precluding within‑subject comparisons. Despite these challenges, the rCPT closely mirrors the structure and attentional demands of the human CPT, reinforcing its translational utility for assessing vigilance in animal models.

In conclusion, our findings demonstrate that vigilance recovery trajectories are agent-specific. The rodent Continuous Performance Task (rCPT) successfully detected delayed vigilance recovery following administration of anesthetic or sedative agents, underscoring its value as a robust translational tool for investigating anesthetic-induced vigilance impairment. Overall, these results highlight the capability of touchscreen-based paradigms to detect subtle cognitive changes, offering insight that may ultimately advance our understanding of postoperative neurocognitive disorders and allow for the development of novel therapeutics.

Supplementary Material

List of supplementary digital content file(s):

1. Supplemental digital content

Abbreviations:

rCPT

rodent continuous performance test

RORR

return of righting reflex

ITI

inter-trial interval

ROC

recovery of consciousness

ANOVA

analysis of variance

5-CSRTT

five‑choice serial reaction time task

PAL

paired-associates learning

MAC

minimum alveolar concentration

PVT

psychomotor vigilance task

EEG

electroencephalography

AEP

auditory evoked potential