Cognitive function following surgery for pelvic organ prolapse

Abstract

Introduction and Hypothesis:

Older women are at higher risk for cognitive dysfunction following surgery. We hypothesized that for women undergoing pelvic organ prolapse surgery, memory function would not be significantly different at delayed postoperative assessment compared to baseline.

Objective:

We sought to compare performance on tests of various neurocognitive domains before and after surgery for pelvic organ prolapse.

Methods:

A prospective cohort study was conducted with women, aged 60 years and older who were undergoing surgery for pelvic organ prolapse. A battery of highly sensitive neurocognitive tests was administered preoperatively (baseline), on postoperative day one (postoperative visit 1), and at the first postoperative clinic visit 4-6 weeks after surgery (postoperative visit 2). The test battery included the scene-encoding memory task, the n-back task, the Iowa gambling task, the balloon analogue risk task, and the psychomotor vigilance task. These tests assessed the neurocognitive subdomains of episodic memory, working memory, decision-making, risk taking, and sustained attention. Two score comparisons were made: between baseline and postoperative visit 1, and between baseline and postoperative visit 2.

Results:

In 29 women, performance on the scene-encoding memory task was worse at postoperative visit 1 than at baseline (2.22 ± 0.4 vs. 2.45 ± 0.6, p < .05) but was better than baseline at postoperative visit 2 (2.7 ± 0.7 vs. 2.45 ± 0.6, p < 0.05). Similarly, performance on the psychomotor vigilance test was worse at postoperative visit 1 than at baseline (p < .001) but there was no difference at postoperative visit 2. There was no difference in performance on the Iowa gambling test, n-back test and balloon analogue risk tasks between baseline and any postoperative visit.

Conclusion:

Cognitive test scores did not worsen significantly between baseline and delayed postoperative assessments in older women undergoing surgery for pelvic organ prolapse.

Introduction

Prior studies indicate that surgery and anesthesia may have deleterious effects on cognitive function in older adults.^1,2 In a landmark study² of over 1200 patients undergoing abdominal, non-cardiac thoracic, or orthopedic surgery, a battery of neuropsychological tests was administered prior to and following the procedure. The authors found that 10% of patients showed signs of persistent cognitive dysfunction at three months postoperatively. Subsequently, reports that anesthesia can adversely affect cognitive function were widely reported in the lay press and have caused considerable alarm in older women who are recommended elective surgery for pelvic organ prolapse.

Women with pelvic floor disorders are likely at higher risk for cognitive dysfunction because they are older and have a high rate of co-existing urinary incontinence, which is known to be associated with cognitive dysfunction.³ A study of the urogynecology patient population found significant rates of baseline cognitive dysfunction that range from 5.3% at ages 65-74 to 30% for patients 85 and older.⁴ In addition to the risks associated with undergoing surgery at an older age, surgery for pelvic organ prolapse (POP) in particular can be of long duration, further increasing potential risk for cognitive dysfunction.⁵ In a group of women with a high rate of baseline cognitive dysfunction (15.7%), Brander et al reported an increased risk of cognitive dysfunction in older women undergoing prolapse and incontinence surgery.⁶ Ackenbom et al also reported an increase in postoperative cognitive dysfunction at two weeks following urogynecological surgery⁷. However, there is relative lack of data on long-term cognitive function in older women undergoing POP surgery who do not have cognitive dysfunction at baseline.

We sought to assess cognitive function before and after surgery for POP in a group of women with no cognitive dysfunction at baseline using sensitive tests of neurocognitive domains, and to consider the domains individually. Our hypothesis was that in women undergoing POP surgery, memory function would not be significantly different at delayed postoperative assessment compared to baseline.

Materials and Methods

We conducted a prospective cohort study of women 60 years of age and older undergoing surgery for POP. Patients were recruited between September 2020 and November of 2021 from urogynecology practices. This study protocol was approved by the Institutional Review Board of our institution (No. XXXXXX). Exclusion criteria included pre-existing cognitive dysfunction as determined by a Mini-Mental State Exam (MMSE) score of less than 24, non-English speaking, inability to read English, severe visual or auditory disorder, and current alcoholism or drug dependence. The MMSE⁸ is a paper-based test with a maximum score of 30, with lower scores indicating worse cognitive function. A cut point of 24 has a sensitivity and specificity of 85% and 90% respectively for detecting dementia.⁹

At the baseline visit, we collected basic demographic information and medical history including cardiovascular and neurologic disorder history. Surgical characteristics were obtained from the medical record, including information regarding route and duration of surgery, type and duration of anesthesia, and any complications. Additionally, pain score was assessed at postoperative visit 1 with a numerical pain rating scale (0-10).

Questionnaires

A series of questionnaires to measure urinary symptoms, anxiety and depression, functional status, frailty, and sleep were administered to each subject at the baseline visit. These included the Urinary Distress Inventory short form (UDI-6), the Hospital Anxiety and Depression Scale (HADS), the Late Life Function and Disability Measure (LLFDI) and the Fried Frailty Index. The UDI-6¹⁰ is a validated 6-item questionnaire that assesses patient distress regarding urinary symptoms. Scores are normalized to a scale of 0-100 with higher scores indicating worse symptoms. The HADS is a validated 14-item self-administered questionnaire developed to assess the severity of anxiety and depression in a hospital outpatient clinic setting. The questionnaire includes seven questions regarding anxiety (anxiety sub-scale) and seven regarding depression (depression sub-scale). Each question is scored on a four-point scale (0-3) and thus each subscale has a score range of 0-21.¹¹ Subscale scores of 11 or higher are considered abnormal. The LLFDI is a validated questionnaire assessing meaningful change in function and disability.¹² We used a validated short form¹³ (15 items) of the physical function instrument, with scores ranging from 15-75 and higher scores indicating less difficulty with activities. The Fried Frailty Index¹⁴ is a validated instrument used to assess frailty in community-dwelling older adults.

Two sleep questionnaires were administered: The Pittsburg sleep quality index (PSQI) and the Epworth sleepiness scale (ESS). The PSQI¹⁵ is a validated questionnaire to assess sleep quality and disturbances over the past one-month time period. Scores range from 0-21, with scores greater than 5 suggesting poor sleep. The ESS is a validated 8-item questionnaire that assesses daytime sleepiness by asking subjects to rate on a scale of 0-3 how likely they are to doze in specific scenarios. Scores range from 0-24, with scores greater than 10 indicating excessive daytime sleepiness.

Cognitive Tests

Participants completed a battery of computer-based neurocognitive tests at three time points: at the preoperative visit (baseline), on postoperative day one (postoperative visit 1, POV1) and at the first postoperative clinic visit (postoperative visit 2, POV2). A prior study has shown that computerized neuropsychological tests are more sensitive than conventional tests for detecting cognitive decline after surgery.¹⁶ The battery included five tests assessing various neurocognitive subdomains: a scene-encoding task¹⁷ (episodic memory), an n-back task¹⁸ (working memory), the balloon analog risk task¹⁹ (BART, assesses risk taking/decision-making), the Iowa gambling task²⁰ (IGT, assesses decision making) and the psychomotor vigilance test²¹ (PVT, assesses sustained attention). The test battery took approximately one hour to complete at each visit. Descriptions of the neurocognitive domains and sub-domains are listed in Table 1.

Table 1:

Description of neurocognitive domains assessed

Neurocognitive domain	Sub-domain	Definition	Real-world example	Test used
Learning and memory	Episodic memory	Learning, storing and retrieving information about personal experiences in daily life, including information about time, place and detailed information about an event itself	Recalling the events, locale, and setting of a college graduation day	Scene-encoding memory task
Executive function	Working memory	Holding information in mind and mentally working with it	Mentally adding up a list of numbers	N-back task
	Decision-making	Selecting an option among a set of alternatives expected to yield different results	A jury weighing evidence to reach a verdict	Iowa gambling task
	Risk-taking/Decision-making	Performing a behavior that carries a chance of potential harm	Investing money in the stock market	Balloon analog risk task
Complex attention	Sustained attention	Prioritizing relevant information that matches a task goal	Maintaining focus on the road while driving a long distance	Psychomotor vigilance task

The scene-encoding task¹⁷ is a test in which subjects are presented with 72 complex visual scenes. Following the initial presentation of scenes, subjects are shown 40 images and asked to recall if they had previously seen the image. Three scores are calculated: hit rate (rate of correctly identifying a previously seen image), false alarm rate (rate of incorrectly indicating that an image had been seen previously), and d’ (a calculated measure of signal detection accuracy). Higher d’ score indicates better function in the episodic memory domain. The primary outcome was d’ score.

For the N-back task¹⁸, subjects are shown a sequence of letters and are instructed to respond by pressing a button whenever a specified “target” appears. Targets are any letter that repeats with a pre-specified number of intervening non-identical letters (e.g. a 1-back test target would be A-A but not A-F-A or A-Q-G-A; a 2-back test target would be A-F-A but not A-A-F or A-C-R-A). The outcomes measured were accuracy rate and mean response time. Higher accuracy rate and lower mean response time indicates better performance in the working memory domain.

In the balloon analog risk task¹⁹, subjects are presented with a series of 60 balloon trials during which they can win or lose potential earnings. On each trial, the subject clicks on the pump to inflate the balloon, with each pump increasing the balloon’s value. Each balloon is programmed to pop at a certain number of pumps but subjects are not informed of the breakpoint of an individual balloon. At any point during each trial, subjects can stop pumping the balloon and collect the money accumulated from that balloon. If the balloon pops, the money is lost for that trial. The outcome is adjusted average pumps, i.e. number of pumps for balloons which did not pop. A higher number of adjusted pumps indicates more risk-taking behavior.

For the Iowa gambling task,²⁰ four decks of cards are displayed on a computer screen. The object is to win money and when a card is drawn, the result is either to add to or subtract from to the subject’s funds during the game. Decks 1 and 2 reward more in the short term but in the long term result in higher losses (considered the disadvantageous decks). Decks 3 and 4 reward less in the short term but bring higher reward in the long term (considered the advantageous decks). Subjects are not told which decks are advantageous. The outcome measured is the number of selections from advantageous decks minus the number from disadvantageous decks (“net score”). A more positive net score indicates better performance on the executive function (decision making) domain.

In the psychomotor vigilance task,²¹ the subject is instructed to respond every time a stimulus appears on the screen but not to respond too soon (“false start”). This test measures the ability to maintain attention and respond in a timely manner to salient signals. The test avoids the confounds of inter-subject (aptitude) and intra-subject (learning) sources of variance.²² Outcomes measured are the number of lapses and mean response time. Lower number of lapses and lower response time indicate better function in the attention domain.

Tests were conducted using MATLAB version 7.12.0, Presentation version 18.3 and PennPVT proprietary software on a Dell Latitude E7240 laptop with an Intel® Core ™ i5-4310U CPU.

We estimated that we would need 22 subjects to detect a significant difference in the d’ outcome of the scene encoding task (primary outcome) at 90% power and alpha of 0.05 before and after surgery. A recruitment target of 27 was set to account for an approximately 20% loss to follow up.

Statistical Analysis

Normality of data was assessed using the Shapiro Wilk test. Demographic and clinical data were described using mean or medians as appropriate. Paired t-test or paired-sample Wilcoxon signed rank test were used as appropriate to make two comparisons: 1) baseline scores with POV1 scores and 2) baseline scores with POV2 scores. Multivariable linear regression was used to assess the impact of patient characteristics on change in cognitive test scores. All analyses were conducted with Stata Statistical Software: Release 16 (College Station, TX: StataCorp LLC).

Results

Sixty patients were approached for enrollment. Thirty declined to participate due to time constraints (25 subjects), lack of interest in participating in a research study (4 subjects), or concerns regarding ability to use technology (1 subject). Thirty patients were consented and completed a baseline visit. Twenty-nine subjects were included in the final analysis and two had partially incomplete data (Figure 1). Baseline demographic and clinical characteristics are presented in Table 2. Mean age was 69 years and mean BMI was 27 kg/m². Most of the subjects were white, not frail, and had some college education. The cohort had a high baseline median MMSE score of 30 (out of a maximum of 30). Overall subjects did not have significant anxiety, depression or urinary distress symptoms. Sleep quality was poor, as measured by the PSQI, but subjects did not have excessive daytime sleepiness, as measured by the ESS.

Figure 1:

Flow diagram of study participants

Table 2:

Demographic and clinical characteristics (n=29)

Age (years)	69.3 (±5.1)
BMI (kg/m2)	26.9 (± 4.6)
Race, n(%)
White	26 (89.7%)
Black or African-American	1 (3.4%)
Asian	1 (3.4%)
Unknown/Not Reported	1 (3.4%)
Education, n(%)
Less than high school	0 (0%)
High school/GED	9 (31.0%)
Associate college degree	2 (6.9%)
Four year college degree	9 (3.1%)
Graduate degree	8 (27.6%)
Unknown/Not Reported	1 (3.4%)
Parity, median (IQR1)	2 (2-3)
Medical co-morbidities, n(%)
Anxiety or panic disorder	3 (10.3%)
Depression	0 (0%)
Stroke or transient ischemic attack	2 (6.9%)
Diabetes (type I or II)	1 (3.4%)
Congestive heart failure	1 (3.4%)
History of myocardial infarction	1 (3.4%)
Cognitive function (MMSE)2 score, median (IQR)	30 (29-30)
Urinary symptom score (UDI-6)3 score	30.8 (±21.0)
HADS4 score, median (IQR)
Depression subscale	2 (0-3)
Anxiety subscale	2 (1-6)
Baseline functional status score (LLFDI) 5	67.9 (± 7.6)
Fried Frailty Index, n(%)
Not frail	27 (93.1%)
Frail	1 (3.4%)
Sleep quality (PSQI)6 score	5.7 (± 3.2)
Daytime sleepiness score (ESS)7, median (IQR)	5 (3-8)
Pain score at postoperative visit 1, median (IQR)	2 (1-4)

^1.Inter-quartile range

^2.Mini-Mental State Exam, range 0-30

^3.Urinary Distress Inventory, Short Form, range 0-100

^4.Hospital Anxiety and Depression Scale, range of each subscale 0-21

^5.Late Life Function and Disability Index, range 15-75

^6.Pittsburg Sleep Quality Index, range 0-21

^7.Epworth Sleepiness Scale, range 0-24

Surgical characteristics of the cohort are presented in Table 3. Most procedures were performed via transvaginal approach. Most subjects were ASA class II and all patients received combination intravenous and inhalation anesthesia. The mean time from surgery date to POV2 was 36.6 days (± 9.6).

Table 3:

Surgical characteristics (n=29)

ASA class, n(%)
I	2 (6.9)
II	20 (68.9)
III	7 (24.1)
IV	0 (0)
Prolapse stage, n(%)
1	0 (0)
2	2 (6.9)
3	26 (89.7)
4	1 (3.4)
Route of surgery * , n(%)
Vaginal	14 (48.3)
Robotic-assisted	8 (27.6)
Abdominal, open approach	7 (24.1)
Intra-operative variables, median(IQR)
Duration of surgery (min)	198 (150-252)
Duration of anesthesia (min)	259 (183-303)
Estimated blood loss (ml)	100 (100-150)

^*Reflects route of surgery for the primary prolapse repair portion of the procedure

Table 4 shows cognitive performance across visits. Compared to baseline, performance on the scene-encoding task as assessed by d’ rate (primary outcome) and hit rate was worse at POV1 but no difference was noted at POV2. False alarm rate did not change significantly from baseline to POV1. A graphical depiction of scene-encoding task outcomes is shown in Figure 2. Performance on the n-back and BART tasks did not change significantly between baseline and POV1. Performance on the IGT improved significantly from baseline to POV1. Performance on the PVT worsened significantly from baseline to POV1 but no difference was noted at POV2.

Table 4:

Performance on neurocognitive test battery

Test and outcome measures	Neurocognitive sub-domain	Pre-op visit score	Post-op visit 1 score	Pre-op vs. post-op visit 1 (p-value)	Post-op visit 2 score	Pre-op vs. post-op visit 2 (p-value)
Scene encoding task	Episodic memory
D’ rate↑		2.45 (±0.6)1	2.22 (±0.5)	0.01 *	2.70 (±0.7)1	0.03 *
Hit rate↑		0.82 (0.8-0.9)2	0.78 (0.7-0.9)2	<0.01 †	0.87 (0.8-0.9)2	0.44†
False alarm rate↓		0.08 (±0.05)1	0.07 (±0.06)1	0.19*	0.04 (0.0-0.1)2	0.04 †
N-back	Working memory
1-back accuracy rate↑		0.96 (0.9-1.0)2	0.96 (0.9-1.0)2	0.53†	0.96 (0.95-1.0)2	0.49†
1-back mean response time (ms)↓		588 (506-675)2	578 (524-630)2	0.48†	560 (498-615)2	0.19
2-back accuracy rate↑		0.92 (0.9-1.0)2	0.95 (0.9-1.0)2	0.46†	0.97 (0.9-1)2	0.02 †
2-back mean response time (ms)↓		775 (697-946)2	807 (629-962)2	0.91†	719 (606-781)2	0.02 *
Iowa Gambling Task	Decision making
Net score↑		−4 (−14-0)2	1 (−10-17)2	0.01 †	−1 (−15-6)2	0.22†
Balloon Analogue Risk Task	Risk taking
Number of adjusted pumps		8.49 (± 2.9)1	8.15 (±3.0)1	0.54*	8.12 (±6.3)1	0.64*
Psychomotor Vigilance Test	Sustained attention
Median reaction time (ms)↓		311 (287-335)2	332 (316-377)2	<0.0001 †	317 (±34)1	0.59*
Number of lapses↓		3 (1-5)2	6 (2-10)2	<0.01 †	2 (1-5)2	0.31†

Bold values indicate statistically significant differences

^↑Higher score for this outcome indicates better cognitive function in this domain

^↓Higher score for this outcome indicates worse cognitive function in this domain

¹Mean (standard deviation)

²Median (IQR)

^*Paired t-test

^†Wilcoxon signed rank test

Figure 2:

Scene-encoding memory task scores across visits

We next assessed potential predictors for decline in performance of the four neurocognitive outcomes that showed worse performance at POV1: d’ and hit rate of the scene encoding task and reaction time and number of lapses of the PVT. Age and MMSE score were not predictive of decline in d’ score (p=0.833 and p=0.369 respectively) and hit rate (p = 0.282 and p= 0.440, respectively). On univariate linear regression, higher pain score at POV1 was predictive of greater decline in d’ score between baseline and POV1 (p=0.016). This persisted after controlling for age and baseline d’ score (p=0.035). For hit rate, higher pain score was marginally predictive of greater decline in score (p=0.05). This relationship did not persist when controlling for age and baseline hit rate (p=0.272). For the PVT outcomes (reaction time and number of lapses), age, pain and MMSE score were not predictive of decline in reaction time (p=0.814, p = 0.982, and p = 0.662, respectively) or number of lapses (p=0.837, p=0.843, and p=0.631, respectively).

Discussion

In our population of older women undergoing POP surgery, we found that the memory and attention domains of cognitive function worsened in the immediate postoperative period but recovered by the first postoperative clinic visit at 4-6 weeks after surgery. Specifically, patients performed worse on the scene encoding task and psychomotor vigilance test at postoperative day 1. However, executive function domains did not exhibit a similar decline in the immediate postoperative period. In fact, performance on the IGT, a measure of the decision-making sub-domain of executive function, improved between the preoperative visit and postoperative visit 1. Taken together, these findings suggest that the cognitive functions most affected in the postoperative period are attention and memory which decline in the immediate postoperative period but had recovered by four weeks postoperatively. Our results are overall reassuring and suggest that while there may be a decline in some aspects of cognitive function in the immediate postoperative time period, cognitive function likely recovers by four weeks postoperatively.

The magnitude of change of computerized neurocognitive tests that is clinically important has not been defined. In our study, we compared change in performance of the cognitive test over time with each patient serving as her own control. The fact that we identified a decline in cognitive function which then recovered to baseline suggests our testing was sensitive enough to detect changes in cognition around the time of surgery. These findings allow us to draw clinically meaningful conclusions. First, we can reassure patients planning to undergo elective pelvic organ prolapse surgery and who do not have cognitive dysfunction preoperatively that cognitive functions such as memory, attention and executive function return to baseline as early as 4-6 weeks after surgery. Next, given that executive functions such as decision making are maintained even at one day after surgery, our findings suggest that patients could provide informed consent for procedures should this be required in the immediate postoperative period. Finally, because sustained attention and memory decline on the first postoperative day, clinicians should recognize that patients may not retain postoperative instructions given prior to discharge. Written instructions and/or counseling of family members or support persons may be useful tools to ensure postoperative instructions are retained.

Our finding that executive function measures did not significantly decline in the immediate postoperative period is consistent with previous work in the anesthesiology literature. Mashour et al²³ conducted a multicenter study of subjects undergoing general anesthesia using a battery of cognitive tests. Similar to our study, these authors found differential recovery of cognitive processes, with executive function processes returning first, and attention and reaction time taking longer to recover. These results are consistent with a “global neuronal workspace” theory, according to which higher order cognitive or consciousness-related functions serve to coordinate lower order cognitive functions and thus may recover first after an insult.²⁴

An important issue in assessment of postoperative cognitive function is whether paper-based or computerized neurocognitive tests should be used. In our study, baseline MMSE score, a commonly used paper based clinical tool to quantify cognitive function, was not predictive of decline on the more sensitive neurocognitive cognitive test performance. This may be because the median MMSE score was high and patients with low MMSE scores were excluded from our study. However, prior studies have shown that the MMSE has limited ability to detect mild cognitive impairment.²⁵ In a cohort of patients aged 70 years and older undergoing pelvic organ prolapse surgery, Brander et al reported a significant increase in cognitive dysfunction using paper-based Mini-COG and clock-drawing tests.⁶ Using a paper-based neurocognitive test battery, Ackenbom et al reported that one-third of patients undergoing prolapse surgery had significant cognitive dysfunction at two weeks after surgery.⁷ These differences may be related to differences in age, rates of baseline cognitive dysfunction, type of tests, and the time periods over which they were administered. In a study examining cognitive changes in 50 patients undergoing coronary bypass surgery, a computerized battery detected all the cases of postoperative cognitive dysfunction identified by the conventional test battery and also five cases that were classified as normal by the conventional tests.¹⁶ Larger studies using both paper-based and computerized tests over different time points are required to determine the optimal tests that should be used to identify cognitive dysfunction following prolapse surgery.

Strengths of this study include its rigorous methodology and use of sensitive computer-based neurocognitive tests. We purposefully excluded patients with cognitive dysfunction at baseline allowing us to isolate the effects of prolapse surgery and anesthesia on cognitive function. Limitations include our small sample size, lack of control group and homogeneous population. Lack of differences between baseline and second postoperative visit scores may also be a result of practice with scores improving due to repeated administration of the test.

Conclusions

Cognitive test scores did not worsen between baseline and the first postoperative clinic visit in women undergoing POP surgery. Further investigation is needed to determine if certain patients are at higher risk of cognitive dysfunction in the postoperative period, and if so, how best to identify these patients preoperatively.

Data availability statement:

data supporting figures 1-2, tables 1-4 are currently not publicly available in order to protect patient privacy, however the data can be made available upon request, pending approval from the University of Pennsylvania institutional review board.