Is the Presence of Mild to Moderate Cognitive Impairment Associated with Self-Report of Non-Cancer Pain? A Cross-Sectional Analysis of a Large Population-Based Study

Abstract

Context

Research, guidelines and experts in the field suggest that persons with cognitive impairment report pain less often and at a lower intensity than those without cognitive impairment. However, this presupposition is derived from research with important limitations, namely inadequate power and lack of multivariate adjustment.

Objectives

We conducted a cross-sectional analysis of the Canadian Study of Health and Aging to evaluate the relationship between cognitive status and pain self-report.

Methods

Cognitive status was assessed using the Modified Mini Mental State Exam. Pain was assessed using a 5-point verbal descriptor scale. For analysis, responses were dichotomized into “no pain” versus “any pain” and “pain at a moderate intensity or higher” versus “pain not at a moderate intensity or higher.” Additional predictors included demographics, physical function, depression, and co-morbidity. Of 5,703 eligible participants, 306 (5.4%) did not meet inclusion criteria, leaving a total of 5,397, of whom 876 (16.2%) were cognitively impaired.

Results

In the unadjusted analysis, significantly more cognitively intact (n=2,541; 56.2%) than cognitively impaired (n=456; 52.1%, P=0.03) participants reported non-cancer pain. There was no significant difference in the proportion of cognitively intact (n=1,623; 35.9%) and impaired (n=329; 37.6%, P=0.36) participants who reported pain at moderate or higher intensity. In multivariate analyses, cognitively impaired participants did not have lower odds of reporting any non-cancer pain (odds ratio [OR]=0.83 [0.68-1.01], P=0.07) or pain at a moderate or higher intensity (OR= 0.95 [0.78-1.16], P=0.62).

Conclusion

Non-cancer pain was equally prevalent in people with and without cognitive impairment, which contrasts with the currently held opinion that that cognitively impaired persons report non-cancer pain less often and at a lower intensity.

Introduction

The prevalence of cognitive impairment is expected to increase as life expectancy in industrialized nations continues to rise, and age represents the leading risk factor for the development of dementia (1,2). It is estimated that 5 million persons currently have cognitive impairment in the United States, and this number will grow to 8 million by 2030 (2). Dementia is a devastating illness and is characterized by gradual cognitive and functional decline with behavioral and psychological symptoms (3).

In addition to cognitive impairment, increasing age is associated with additional morbidities such as arthritis, diabetes, spinal stenosis, vascular disease, and postherpetic neuralgia. Each of these disorders increases the likelihood of older adults experiencing persistent non-cancer pain (4-7). Non-cancer pain results in considerable discomfort, as well as a physical, emotional, and social burden for persons with and without cognitive impairment (8). That is, non-cancer pain is associated with worse health outcomes, including functional impairment, depression, decreased appetite, impaired sleep, and poorer self-reported health (8,9).

Studies of both cognitively intact and impaired populations report more than 50% of community-dwelling older adults live with non-cancer pain (10-12). Nonetheless, research suggests and guidelines endorse that persons with cognitive impairment report pain less often, and at a lower intensity, than do those without cognitive impairment (13-18). Reasons cited for this finding are that declines in memory, language, and executive function result in lower levels of reported pain in those with impairment (13). Also, parenchymal brain changes that lead to cognitive impairment may modify the perception of pain by altering signal processing within the amygdala, decreasing the affective contribution of pain, which in turn lessens reported pain intensity (19). Despite the findings of more recent studies that persons with mild to moderate cognitive impairment provide reliable and valid pain reports when appropriate scales are used (8), health care providers risk undervaluing the pain report of persons with cognitive impairment. This may contribute to therapeutic nihilism.

This perception of a decrease in pain self-report in persons with cognitive impairment is concerning since the literature from which this conclusion is drawn contains important, unrecognized limitations. For example, several published studies have small sample sizes, resulting in inadequate power to provide sufficient evidence that persons with cognitive impairment report less pain (16-18). The majority of available studies also lack adjustment for variables know to be associated with pain self-report, such as co-morbidity and depression (15). At the same time, studies demonstrate that cognitive impairment is a risk factor for pain under-treatment, which is not only associated with physical discomfort but also higher rates of delirium, depression, and functional impairment (8,20-23).

The purpose of this large, nationwide, population-based study of older adults is to determine whether or not cognitive impairment is truly associated with a lower frequency and intensity of self-reported non-cancer pain. The current study provides sufficient power to test the hypothesis that those with cognitive impairment self-report less pain when controlling for important confounding covariates.

Methods

Participants

This is a cross-sectional analysis of the Canadian Study of Health and Aging (CSHA). The CSHA was a national longitudinal study designed to provide information about the epidemiology of dementia and its impact on caregivers; it has been previously described in detail (24-25). Briefly, people aged 65 years or over were recruited with representative samples drawn from 36 cities and their surrounding rural areas across Canada, including people dwelling in the community and those in institutions. Participants were initially assessed in 1991, with subsequent interviews occurring at 5-year intervals for a total of three waves (1991, 1996, and 2001).

The current study includes community-dwelling older adults who completed the screening questionnaire from the second wave (1996, at which time the youngest participants would have been aged 70 years) which was the first and only time the survey incorporated a measure of self-reported pain. Eligible participants were excluded if they had greater than moderate cognitive impairment, as this may impact the reliability of the pain question, and did not report having cancer in the past year (to minimize inclusion of participants with cancer-related pain). All participants provided written informed consent and the protocol was approved at each participating institution. These secondary analyses have been reviewed and approved by the University of Chicago's IRB.

Predictor and Outcome

All persons who agreed to participate in the study completed a screening questionnaire conducted by trained research personnel in the person's home. Cognitive status (predictor) was assessed using the Modified Mini Mental State Exam (3MS). The 3MS builds upon the Folstein Mini-Mental Exam by adding four tasks (date and place of birth, animal naming for categorical word retrieval, similarities of words for abstract and conceptual thinking, and a second delayed recall) and allowing partial credit on some items. The 3MS has an expanded range from 0 to 100 with lower scores indicating greater cognitive impairment. It has been validated in a variety of community-dwelling populations, and a cutoff of 77 or less is sensitive and specific for identifying cognitive impairment (26,27).

Non-cancer pain (outcome) was assessed using a 5-point verbal descriptor scale (VDS), which measures the presence and intensity of pain. The validity and reliability of this measure have been established in both cognitively intact and impaired populations (28-31). Participants were asked, “How much bodily pain have you had during the past 4 weeks?” Response categories were: 1 = none, 2 = very mild, 3 = moderate, 4 = severe, and 5 = very severe. The outcome was dichotomized for analysis two ways to examine whether pain report varied with cognitive status: 1) the presence of non-cancer pain (any pain) and 2) severity of non-cancer pain (pain at a moderate intensity or higher).

Covariates

Measures that may modify the relationship between cognitive status and pain presence were also included in the analysis. Functional status was examined using the Older Americans Resources and Services Instrumental Activities of Daily Living (IADL) and Activities of Daily Living (ADL), measures which have been validated in persons with and without cognitive impairment (32,33). Each of the five IADL (getting to distant places, using the telephone, going shopping, preparing own meals, and doing housework) and seven ADL (eating, dressing, personal care, walking, getting out of bed, taking a bath, and using the toilet) measures were dichotomized into any impairment versus no impairment. Depression was examined by using a 5-item mental health screening questionnaire (34). Each of the five questions assess mental health (feelings of anxiety and depression) on a 6-point scale ranging from “none of the time” to “all of the time,” where higher scores indicate more symptoms of depression and total scores range from 5 to 30. To facilitate the analyses, participants designated as having depressed mood were those whose summary scores were in the top 25% of all participant scores. Co-morbidities were assessed by asking participants about conditions present in the past year from 15 body systems, and each was dichotomized as “yes” or “no.” The body systems included high blood pressure, heart or circulation problems, arthritis or rheumatism, Parkinson's disease or other neurologic problem, eye trouble, ear trouble, chest problems, trouble with stomach or digestive system, back problems, bladder control problems, problems controlling bowels, fractures, cancer, diabetes, and foot problems. Affirmative responses were summed to create an overall co-morbidity index ranging from 0-15, with higher scores indicating greater morbidity. Self-reported health was assessed using a 5-point scale ranging from poor to excellent health. Demographic information included age, gender, marital status, race/ethnicity, and education.

Statistical Analysis

We conducted a prospective power analysis to determine the feasibility of testing for differences in pain prevalence between cognitively intact and impaired participants. We assumed a prevalence of cognitive impairment and non-cancer pain of 20% and 50%, respectively (2,3,8). For a sample size of 4,800, approximately 800 persons would have cognitive impairment and 4,000 would not, with 400 and 2,000 participants with and without non-cancer pain, respectively. These sample sizes have 80% power to detect a difference in pain self-report of 10%, which is thought to represent a clinically significant difference, assuming a pain prevalence of 60% in cognitively intact and 50% in cognitively impaired persons (60% - 50%). This power calculation assumes a two-tailed test and a Type I error rate of 5%. Descriptive statistics were analyzed by categorizing cognitive status (predictor) into two groups. Cognitively intact participants had 3MS scores above 77, whereas cognitively impaired participants had scores between 50 and 77 (26,27). Continuous variables were summarized as means and standard deviations, and categorical variables as percentages. T-tests were used to compare continuous variables and chi-square was used for dichotomous variables. The outcome variable of interest, non-cancer pain, was classified in two ways: 1) any non-cancer pain report, and 2) pain at a moderate intensity or higher. The first measure assesses pain presence, and the second pain intensity. Associations between the outcome non-cancer pain and predictor cognitive status were compared using Chi-square analysis. The variables considered for multivariate analysis include age, gender, race, education, marital status, education, IADL impairment, ADL impairment, depressed mood, co-morbidity, and self-reported health. Any differences between exposure groups with a P<0.10 or previously shown or thought to be associated with the outcome of interest (i.e., non-cancer pain) were included in multivariate analysis. If statistical adjustment was necessary to correct for baseline differences, these factors were also included in the multivariate model. We used logistic regression to evaluate the two models of non-cancer pain (presence of pain and pain at a moderate intensity or greater). We computed a correlation matrix that included all the independent variables to assess for multicollinearity. For variables found to be highly inter-correlated, we based inclusion in the final models upon perceived clinical relevance. Marital status, ADL impairment, and self-reported health were not included in the final multivariate models due to multicollinearity. Marital status was correlated with age (r = -0.30) and male gender (r = 0.43), so it was left out of the multivariate models. ADL and IADL impairment were correlated (r = 0.44). IADL impairment was selected for inclusion in the final multivariate models because it was more common in the sample population and is also a better indicator of whether someone can live independently. Self-reported health was removed from the multivariate models because it was correlated with IADL impairment (r=0.29), depression (r=0.33), and co-morbidity (r=0.43), indicating it represented a multidimensional measure. Statistical analyses were performed using commercially available software (SPSS; SPSS Inc, Version 16.0, Chicago, IL). Funding sources had no role in the design, analysis, and interpretation of study results.

Results

Of 5,703 community-dwelling older adults who completed the second wave of the CSHA, a total of 306 were excluded: 103 were missing a pain response (1.8%), 87 had a 3MS score less than 50 (1.5%), and 116 had cancer diagnosed in the past year (2.0%). This left 5,397, or 94.6% of the original sample. A total of 876/5397 or 16.2% met this study's criterion for cognitive impairment. A comparison of patient characteristics by cognitive status (intact versus impaired) is displayed in Table 1. Persons with cognitive impairment tended to be older, more physically impaired, and to have higher depression scores.

Table 1.

Sociodemographics and Clinical Characteristics of the Second Wave of the Canadian Study of Health and Aging by Cognitive Status, n=5,397

Characteristic	Cognitively Intact n=4,521 (83.8%)a	Cognitively Impaired n= 876 (16.2%)a	P-value
Age, yrs (mean, SD)	79.2 (5.9)	84.0 (6.8)	<0.001
Gender (% female)	61.1	57.0	0.02
Race (% Caucasian)	98.9	98.7	0.60
Education (% high school or less)	71.8	88.3	<0.001
Marital status (% married)	48.5	37.4	<0.001
Instrumental activity of daily living (% any impairment)	39.7	63.7	<0.001
Activity of daily living (% any impairment)	17.2	36.3	<0.001
Depressed mood (% depressed)b	27.6	43.2	<0.001
Co-morbidities (total number in 15 possible categories, median)	3.7	3.9	<0.001
Self-reported health (% very good or good)	83.9	68.2	<0.001

SD = standard deviation.

^aCognitively intact includes participants with Modified Mini Mental State Exam scores higher than 77, whereas cognitively impaired Modified Mini Mental State Exam scores range from 50-77.

^bDepressed mood is defined as being in the top 25% of total scores on the mental health screening questionnaire among participants of the Canadian Study of Health and Aging, where increasing values mean greater depression.

Table 2 displays cognitively intact and impaired persons’ reports of pain on the VDS. Table 3 displays the unadjusted results examining the association between cognitive status and non-cancer pain. Cognitively-intact participants were significantly more likely to report non-cancer pain, 2,541 (56.2%), than were impaired participants, 456 (52.1%), P=0.03. Conversely, there was not an association between a participant's cognitive status and non-cancer self-report of pain at a moderate intensity or higher. Among cognitively intact participants, 1,623 (35.9%) reported pain at a moderate intensity or higher, whereas among cognitively impaired participants 329 (37.6%) reported pain at that level, P=0.36.

Table 2.

Participant Response to the 5-point Verbal Descriptor Pain Scale by Cognitive Status, n=5,397

Pain Intensity (0-4)	Cognitively Intact (n=4,521) n (%)	Cognitively Impaired (n=876) n (%)
No pain (0)	1980 (43.8%)	420 (47.9%)
Very mild pain (1)	918 (20.3%)	127 (14.5%)
Moderate pain (2)	1107 (24.5%)	203 (23.2%)
Severe pain (3)	418 (9.2%)	95 (10.8%)
Very severe pain (4)	98 (2.2%)	31 (3.5%)

Table 3.

Relationship Between Pain Self-Report and Cognitive Status, Non-Adjusted, n=5,397

	Cognitively Intacta	Cognitively Impaireda	P-value
Any painb
Yes	2541/4521 56.2%	456/876 52.1 %	Chi-square=5.12 P= 0.026
No	1980/4521 43.8 %	420/876 47.9 %
Pain at moderate intensity of higherb
Yes	1623/4521 35.9 %	329/876 37.6 %	Chi-square=0.87 P= 0.357
No	2898/4521 64.1 %	547/876 62.4 %

^aCognitively intact includes participants with Modified Mini Mental State Exam scores higher than 77, whereas cognitively impaired scores range from 50-77.

^bSelf-report of pain dichotomized from categorical values of none, very mild, moderate, severe, very severe.

Table 4 displays the multivariate regression analysis for the outcome non-cancer pain, any versus none, by the relevant predictors including cognitive status. Cognitively impaired participants did not have a statistically significant decreased odds of reporting non-cancer pain compared to those who were cognitively intact, OR=0.83 (0.68-1.013), P=0.07. The logistic regression had a Hosmer-Lemeshow Chi-square of 2731.84 (P=0.23) and a pseudo R² of 0.11 (P<0.01) both indicating a satisfactory goodness of fit for the model. Table 5 displays the multivariate regression analysis for the outcome non-cancer pain at a moderate intensity or higher, yes versus no, by relevant predictors including cognitive status. Again, cognitively impaired participants did not have a significantly decreased odds of reporting pain at a moderate intensity or higher compared to those who were cognitively intact, OR= 0.95 (0.78-1.16), P=0.62. The logistic regression had a Hosmer-Lemeshow Chi-square of 2668.32 (P=0.54) and a pseudo R² of 0.12 (P<0.01), both indicating a satisfactory goodness of fit for the model. For each logistic regression, the predictors associated with pain self-report were similar and included age, female gender, impairment in IADL's, depressed mood, and greater co-morbidity.

Table 4.

Logistic Regression of the Presence of Non-Cancer Pain by Patient Demographics, Cognitive Status, and Variables that Add to the Model, n=5,397^a

Characteristic	No Pain versus Any Pain
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Age	1.00 (0.99-1.01)	1.02 (1.01-1.03)
Male gender	0.65 (0.55-0.72)	0.77 (0.67-0.89)
Caucasian race	0.96 (0.58-1.56)	0.85 (0.46-1.55)
Education
Greater than high school	Reference	Reference
High school	0.98 (0.86-1.11)	1.02 (0.87-1.19)
Grade school	1.03 (0.88-1.19)	1.16 (0.97-1.37)
Instrumental activity if daily living, impaired	1.85 (1.65-2.07)	1.32 (1.15-1.53)
Depressed mood	2.37 (2.09-2.69)	1.69 (1.46-1.96)
Number of co-morbidities	1.45 (1.41-1.49)	1.41 (1.36,1.46)
Cognitively impairedb	0.85 (0.73-0.98)	0.83 (0.68-1.01)

CI = confidence interval; OR = odds ratio.

^aLogistic regression was used to develop the most parsimonious model of pain presence with patient demographics, cognitive status, and other variables that add to the model and those thought to be clinically relevant.

^bCognitive impairment defined as a Modified Mini Mental State Exam score between 50 and 77.

Table 5.

Logistic Regression of Non-Cancer Pain at Moderate Severity or Higher by Patient Demographics, Cognitive Status, and Variables that Add to the Model, n=5,397^a

Characteristic	No Pain versus Moderate Pain Severity or Higher
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Age	0.99 (0.98-0.99)	1.02 (1.01-1.03)
Male gender	0.60 (0.53-0.67)	0.74 (0.64-0.86)
Caucasian race	0.97 (0.59-1.27)	0.95 (0.51-1.72)
Education
Greater than high school	Reference	Reference
High school	0.99 (0.78-1.03)	0.99 (0.84-1.17)
Grade school	0.81 (0.70-0.94)	0.94 (0.78-1.12)
Instrumental activity if daily living, impaired	2.16 (1.92-2.42)	1.45 (1.26-1.68)
Depressed mood	2.27 (2.34-3.00)	1.77 (1.53-2.04)
Number of co-morbidities	1.43 (1.39-1.47)	1.35 (1.31-1.40)
Cognitively impairedb	1.07 (0.93-1.25)	0.95 (0.78-1.16)

CI = confidence interval; OR = odds ratio.

^aLogistic regression was used to develop the most parsimonious model of pain presence with patient demographics, cognitive status, and other variables that add to the model and those thought to be clinically relevant.

^bCognitive impairment defined as a Modified Mini Mental State Exam score between 50 and 77.

Discussion

This study found non-cancer pain to be a prevalent symptom in persons with and without cognitive impairment. The proportion of participants reporting pain at a moderate intensity or higher was over 30% for both cognitively intact and impaired persons. While cognitive impairment was associated with a lower frequency of non-cancer pain report in the unadjusted analysis, this difference was no longer significant with multivariate adjustments. The report of non-cancer pain at a moderate intensity or higher was not associated with the presence of cognitive impairment in the unadjusted or the multivariate analysis. This supports the hypothesis that cognitively impaired persons can reliably report pain. We found the report of non-cancer pain to be associated with poorer health such as functional impairment, depressed mood, and increased co-morbidity. These have been reported in previous studies and supports the validity of our findings (8).

Non-cancer pain remains a frequent complaint among community-dwelling older adults (8). Other population-based studies that have examined the prevalence of pain among older adults have found similar results. For example, a Swedish study found 60% of older adults over the age of 75 reported pain (9). A community-based study in Scotland reported a chronic pain prevalence of 62% (10). Sixty-six percent of older adults in Iowa reported joint pain (35). In comparison to mostly Caucasian populations in the United States, Canada, and Western Europe, little is known about the prevalence of non-cancer pain in older adults in other ethnic groups. The current study does not address this important question.

Previous studies suggest persons with cognitive impairment report non-cancer pain at a lower frequency and intensity compared to cognitively intact persons (15-18). However, available research has important limitations. Two of the studies enrolled 20 or fewer participants with cognitive impairment as the comparison group, diminishing the power of either study's conclusions (16,18). Another study reported that only 35% of the original sample was included in the data analysis, raising questions about the generalizability of the findings (17). The one investigation large enough to support a multivariate analysis did not provide sufficient methodological detail to draw meaningful conclusions about the study's findings (15). A recent study of 126 participants using multivariate adjustments found that persons with cognitive impairment were not less likely to report pain than those without impairment when resting, but impaired persons did report pain less frequently than intact persons when measured after activity (36). However, this analysis did not control for other factors that we found modified bivariate associations with pain self report between these groups, namely co-morbidity, depression, and functional status. In addition, all but one of these studies were primarily conducted in long-term care or assisted living settings, further limiting the generalizability of the results to the large population of community-dwelling older adults with cognitive impairment. Unlike previous studies, the present investigation benefits from a broad data set from a large population-based study with few of the eligible participants being excluded.

The findings of the present study also corroborate laboratory and imaging evaluations of pain response. Persons with cognitive impairment reported similar pain thresholds in the laboratory to those of cognitively intact controls (37). Similarly, a study using functional MRI found no difference in central nervous system processing to mechanical pressure between cognitively intact and impaired persons. Functional imaging demonstrated activity in the medial and lateral pain pathways in both groups; similar pain-related activity was observed in the cingulate, insula, and somatosensory cortices (38). Taken together, it does not appear that pain perception and processing lessens in persons with mild to moderate cognitive impairment which is consistent with the findings of this study. Current guidelines that suggest older adults with cognitive impairment report less pain and at a lower intensity than persons without impairment may be misleading (13,14). This supposition may contribute to known under-treatment of pain in this population.

The finding that there is no difference in the frequency of pain self-report between cognitively intact and impaired persons must be understood in light of previous research. One concern is whether persons with cognitive impairment can reliably self-report pain experience. Self-report is considered the best source for pain assessment in older adults with and without cognitive impairment (8,14). The VDS, the instrument used in this investigation to measure pain, demonstrates concurrent validity and reliability in cognitively intact and mildly to moderately impaired populations (28-31). The ability of persons with cognitive impairment to accurately report their pain experience is supported by this study's finding of no association between the presence of cognitive impairment and the frequency of non-cancer pain self-report. Of note, persons with severe cognitive impairment were not included in the analyses as the reliability of pain self-report in this population has not been established. More detailed information about the participant's pain experience including the type, location, frequency and management of pain were not included in the questionnaire. Another consideration is the use of “four weeks” in the stem of the pain question. While this may seem like a prolonged period of time for someone with cognitive impairment to remember his or her pain experience, the stability of pain reports in cognitively intact and mildly to moderately impaired persons over several weeks has been established (15). Although the data were collected in 1996, we believe the study's findings remain current and relevant as neither the assessment nor treatment of non-cancer pain or cognitive impairment has significantly changed since the data were collected. The fact that few longitudinal studies of aging actually incorporate physical symptom measures underscores the importance of these results. Finally, the findings remain timely as they provide a more complete picture in a large population based study of non-cancer pain in persons with cognitive impairment compared to that which is available in the literature.

In conclusion, in this large, population-based study of community-dwelling older adults, we did not find that the presence of mild to moderate cognitive impairment was associated with decreased frequency or intensity of non-cancer pain self-report. As such, pain reports in persons with cognitive impairment need to be addressed and treated as those in any other patients. This is particularly important as cognitive impairment is a risk factor for pain under-treatment, which has been associated with not only increased physical discomfort but also higher rates of delirium, depression, and functional impairment (8,21-23).