Assess moderating effects of functional decline on the associations between late-life cognitive impairment (CImp) and diabetes and Parkinson's disease (PD); including controls for Mexican-American ethnicity, education, life satisfaction, age, and sex.
In-home interviews with 1,252 elderly Mexican-American (N = 799) and non-Hispanic white (N = 353) residents of El Paso County, Texas. CImp measured by MMSE, CLOXI and CLOXII; functional impairment (ADLimp) as impairment in 1-10 activities of daily living. Our hypothesis is that ethnicity will effect variance of diabetes, hence cognitive decline.
Logistic regression analyses--After controlling for effects of all above-cited variables, PD remains significantly associated with the three measures of CImp, including impairment in executive control function. Controlling for ADLimp does not extinguish the significant association between diabetes and CImp on any of the three measures. However, no significant degree of association between diabetes and CImp remains after other control variables (including Mexican-American ethnicity) have been added to the equation.
1) PD findings are statistically and clinically meaningful. After controlling for all other variables, the OR for respondents diagnosed with PD (compared to their non-diagnosed counterparts) is 1.42 for MMSE impairment (95% CI1.10-15.53); 4.12 for CLOXI impairment (95% CI 1.07-15.85); and 10.51 for CLOXII impairment (95% CI 2.55-43.41). 2) The connection between diabetes and CImp is problematic; our findings suggest that many of the earlier-reported research findings linking diabetes with CImp may be an artifact of other intervening phenomena such as regional and ethnic differentials in prevalence rates for diabetes. 3) The relationship between CImp and ADLimp is strong and clinically meaningful; for each unit increase in ADLimp there is a corresponding 1.33 increase in odds for MMSE impairment (95% CI 1.15-1.55). For impairment on CLOX1 and CLOX2 the ORs are 1.22 (95% CI 1.05-1.42) and 1.21 (95% CI 1.03-1.42). 4) When coupled with other research findings, Mexican-American ethnicity may itself represent a risk factor for CImp. After controlling for effects all other variables, El Paso's elderly Mexican Americans possess odds 2.46 times greater than those for NHWs in MMSE impairment (95% CI 1.42-4.25); 1.53 (95% CI 1.06-2.20) times greater for impairment in executive control function (CLOXI); and 2.35 times greater for impairment in ability to perform a simple copying task (1.35-4.09). 5) Our findings point to the importance of utilizing a number of different screening devices for assessment of cognitive function in order to increase the likelihood that results can be taken as valid, dependable, and clinically meaningful for elderly individuals a Hispanic ethnicity.
Cognitive impairment, Diabetes, Parkinson's disease, Hispanic
A number of studies have linked Parkinson's disease (PD) [1-4] and diabetes [5-11] to cognitive impairment (CImp). CImp has also been connected linearly [12-16] and/or reciprocally [17] to functional decline. However, linkages between PD, diabetes and CImp have either not been universally reported [13,18,19] or have been described as relatively complex [2,20-22]. Diabetes is known to be more common among Hispanic populations and we conducted this study to examine the interactions between ethnicity, and the occurrence of diabetes, Parkinsonism and cognitive decline in a community sample [23].
We report logistic regression analyses findings from a community sample of elderly Mexican Americans and non-Hispanic whites demonstrating that degrees of associations between diabetes and PD with CImp persist after controlling for effects of functional decline. However, after adding effects of a number of control variables (including ethnicity) to the equation, association between diabetes and CImp is no longer significant. We suggest that these findings may well be an artifact of the high prevalence of diabetes among elderly Mexican Americans.
Hypothesis: That ethnicity and diabetic disease will vary in a manner that allows prediction and assessment of cognitive decline and Parkinson's Disease, given the known variable of ethnic differences between Hispanic and non-Hispanic populations.
The purpose of this study was to assess the association between ethnicity, diabetes and Parkinson's disease related cognitive decline in an elderly sample in Hispanic and non-Hispanic individuals.
We interviewed a stratified random sample of 1,152 non-institutionalized elderly (65 years and older) residents of El Paso County, Texas, during 2000-2001. This sample represents 84 percent of respondents originally contacted for interview. The survey instrument was translated into Spanish through cross-translation. In-home interviews were conducted in Spanish or English, depending on respondents' preference. Twenty-four census tracts were randomly selected from 94 census tracts specified in the 1990 census after stratification for median income and ethnic composition. Specifically, eight census tracts were randomly selected from the lower, middle, and upper-middle income categories. Each of these 24 census tracts was screened through telephone interviews to identify households containing one or more members 65 years of age or older. Screening interviews also provided demographic information including respondent's age, gender, and ethnicity. Independent random sampling procedures were then established for Hispanics and non-Hispanic Whites. These procedures ensure virtually equal representation, within each of the three census-tract income categories, of women and men, and age cohorts 65-74 and 74-and-older. Mean and median ages for the entire sample are 74.9 and 74.0. Fifty percent of the sample is female.
CImp has been ascertained with three measures: A global assessment of cognitive function through the MMSE [24]; an assessment of executive control function through CLOXI [25]; and a simple CLOXII copying task that is somewhat independent of executive skills [25]. Executive control functions measured by CLOXI are cybernetic processes that control one's ability to initiate, sequence, and monitor complex goal-directed activities [26]. CLOXII measures constructional praxis [25]. CLOXI and CLOXII have been validated in elderly Mexican-American and non-Hispanic White populations [27]. CLOX has also been shown to improve on the sensitivity of the MMSE to various sociocultural influences [28,29]. Impaired cognitive function is operationally defined as a score less than 25 (MMSE), less than 10 (CLOXI), and less than 12 (CLOXII). These cutoff criteria are based on previously reported performance results from cognitively impaired populations [25].
Specific diseases are assessed by asking whether a doctor had told respondent that he/she had a heart attack (coronary/myocardial infarction/coronary thrombosis), high blood pressure, diabetes (sugar in urine/high blood sugar), stroke, broken hip, cancer, arthritis (rheumatism), or Parkinson's disease. Bivariate cross-tabulation findings indicate that, of these 8 diseases, only diabetes and PD significantly associate with all three measures of cognitive function; only diabetes and PD are included in multivariate analyses. Almost 25% (N = 284) of our sample have been formally diagnosed with diabetes, as opposed to less than 2% with PD (N = 17). Percentages for diabetes among Mexican Americans and NHWs are 30.5 versus 11.7, a significant difference clinically as well as statistically. Percentage differences in PD for the two ethnic groupings are not statistically significant (Mexican Americans = 1.4%; NHWs = 1.7%). As can be seen in the 95% confidence interval summaries for PD (Table 1) the small number of formally diagnosed PD respondents renders odds ratio (OR) interpretation problematic. On the other hand, statistically significant results for small samples are highly salient. The degree of severity of Parkinson's disease was considered to be mild to moderate in this sample. Patients with the diagnosis of Lewy Body dementia were excluded in order to maintain a minimum number of confounding variables.
Table 1: Logistic regression analysis predicting impaired cognitive functiona for parkinson's disease and diabetes mellitus (N = 1152). View Table 1
Functional decline is defined as impairment in activities of daily living (ADLimp). ADLimp is measured by asking respondents if they can perform 10 activities without help (telephoning, drive car/travel alone on a bus or taxi, go grocery shopping, prepare meals, do light housework, take medicine, handle money, do heavy work around the house, walk up-down stairs, walk half-mile).
Control variables include ethnicity, life satisfaction, education, age and sex. Ethnicity is defined through respondent's self-selection and childhood and adolescent developmental history. This procedure has yielded samples of 799 elderly Mexican Americans and 353 NHWs. Life satisfaction is measured with a 12-item scale [30] containing the response format, "Disagree", "?" "Agree". Total scale scores can range 12-36. Chronbach's alpha for this scale is 0.77 among the El Paso County sample. Life Satisfaction is included as a control variable because quality of life and/or life satisfaction appear to be associated with diabetes [31] and PD [2,32]. This variable is also linked with functional capacity [31] and cognitive function [32].
Educational attainment is included because of its powerful influence on time of onset for cognitive impairment [33]; and is measured as years of formal education, including ordinal categories for "high-school graduate", college graduate", and "at least some post-college education". Thirty-five percent of the Mexican-American sample possesses at least a high school degree, thus providing a meaningful comparison between ethnic groupings. Age (65-74 versus 75 and older) and sex (male = 0; female = 1) are also included as control variables.
Logistic regression analyses are conducted for each of the three cognitive function measures. These analyses provide tests of significance and odds ratios for each independent variable (IV) on CImp after controlling for effects of all other IVs in the statistical equation. Analyses are conducted in three blocks. PD and diabetes are entered in Block 1. ADLimp is added to the equation in Block 2; and effects of all 8 independent variables are assessed in Block 3.
Logistic regression results are summarized in Table 1. Shown for MMSE, CLOXI and CLOXII are variable-by-variable effects on CImp after controlling for effects of all other variables in the equation. Blocks 1 in Subtable 1A, Subtable 1B, and Subtable 1C indicate that diabetes and PD are significantly, and somewhat independently, associated with CImp on all three measures. The OR of 1.95 for diabetes in Subtable 1A (after controlling for effects of PD) indicates that respondents with diabetes possess odds for MMSE-impairment 1.95 times those of their non-diabetic counterparts. The corresponding OR for respondents diagnosed with PD is 6.31. Similar outcomes are found in Subtable 1B, and Subtable 1C when CImp is measured with CLOXI and CLOXII. The odds for executive function impairment (CLOXI) among those enduring PD are 6.31 times those for non-sufferers. The corresponding OR for elderly respondents diagnosed for diabetes is 1.47. Both diseases appear to impair cognitive performance related to the straightforward copying task required in CLOXII. Relevant ORs for PD and diabetes in Subtable 1C are 14.74 and 1.87.
ADLimp is added to the equation in Blocks 2 of Subtable 1A, Subtable 1B, and Subtable 1C. In none of the three measures of cognitive function does ADLimp render effects of PD and diabetes on CImp statistically non-significant. ORs for PD and diabetes relating to MMSE impairment are 4.17 and 1.79; for CLOXI impairment (executive control function), 5.07 and 1.40; and for CLOXII impairment, 11.10 and 1.77. All three variables (PD, Diabetes, and ADLimp) independently affect CImp whether this construct is operationally defined by MMSE, CLOXI, or CLOXII. The OR depicted for ADLimp in Block 2 of Subtable 1A indicates that each unit increase in ADLimp elicits a 1.50 unit increase in odds for MMSE impairment; corresponding 1.34 unit increases in ORs (Blocks 2, Subtable 1B, and Subtable 1C) are found for CLOXI and CLOXII impairment.
A clear separation occurs in the relative explanatory powers of PD and diabetes on CImp when all IVs are represented in the statistical equation (Blocks 3, Subtable 1A, Subtable 1B, and Subtable 1C). Remarkably, PD, despite the small numerical size of this diagnosed sample, continues to significantly associate with CImp after controlling for effects of Mexican-American ethnicity, education, life satisfaction, age, and sex. Effects of PD on CImp remain salient whether this phenomenon is defined globally through MMSE (OR = 4.14; Block 3, Subtable 1A), as executive control with CLOX1 (OR = 4.12; Block 3, Subtable 1B), or through CLOXII's simple copying task (OR = 10.51; Block 3, Subtable 1C). Diabetes, however, no longer significantly associates with CImp on any of the three measures.
Like PD, ADLimp continues to significantly associate with CImp after effects of all other IVs are removed from this variable. With respect to MMSE impairment (Block 3, Subtable 1A) the OR for ADLimp is 1.33; for CLOX1 impairment (Block 3, Subtable 1B), 1.22; and for CLOX2 impairment (Block 3, Subtable 1C), 1.21. Nevertheless, findings reported in Blocks 3 of the three sub-tables in Table 1 clearly indicate that ADLimp by itself produces little consequence on either the PD-CImp or the diabetes-CImp relationship.
Finally, it should be noted parenthetically that, in addition to PD and ADLimp, Mexican-American ethnicity and advanced age independently contribute to CImp on all three measures (Blocks 3 or Subtable 1A, Subtable 1B, and Subtable 1C). Conversely, education appears to prolong the onset of CImp, as measured by MMSE and CLOX. Life satisfaction, negatively associates with CImp when this variable is assessed globally (MMSE) or through a simple copying task (CLOXII). However, after controlling for effects of other variables in the equation, life satisfaction does not significantly contribute to impairment of executive function (CLOXII, Block 3, Subtable 1B).
Our findings indicate in a stratified, random sample of 1,152 Mexican-American and NHW elderly residents of El Paso County, Texas, that PD is significantly associated with CImp on three disparate measures of this phenomenon. Despite the small number of respondents formally diagnosed with PD, its association with CImp remains significant after the powerful effects of education, age, ADLimp, and Mexican-American ethnicity are removed from the statistical equation (Blocks 3, Subtable 1A, Subtable 1B, and Subtable 1C). The ORs for PD sufferers are 4.14 on the MMSE, 4.12 on CLOXII, and 10.51 on CLOXII. It should be noted that effects of tremor on ability to draw do not contribute to cognitive function scoring on either of these Clock drawing tests [25], or drawings associated with the MMSE [24]. Thus, along with above cited research findings, we suggest that our PD findings are clinically meaningful. We suggest that medical professionals involved with the treatment of PD patients be aware of the possibility of a PD-CImp connection during ongoing treatment.
The connection between diabetes and CImp is problematic. Results presented in Blocks 2 of Subtable 1A, Subtable 1B, and Subtable 1C indicate that after adjusting for PD and ADLimp, and prior to controlling for effects of Mexican-American ethnicity, educational attainment, life satisfaction, advanced age, and sex, diabetes somewhat independently associates with cognitive impairment on all three measures. However, the inclusion of all IVs in Block 3 of each sub-table renders diabetes a non-significant predictor of CImp, regardless of how this phenomenon is measured. It is possible that the significant findings for diabetes depicted in Blocks 1 & 2 may be an artifact of the relatively high prevalence of diabetes (30.5%) for El Paso's elderly Mexican-American population (compared to an 11.7% prevalence rate for NHWs). In order to test this idea we cross tabulated diabetes and CImp, controlling for Mexican-American-NHW ethnicity. No significant degree of association was found between diabetes and CImp among NHWs on either of the three measures. Diabetes did associate with MMSE impairment (p < 0.05) among elderly Mexican Americans, but no significant findings were found for CLOXI or CLOXII impairment. The MMSE findings are in line with those reported for the Hispanic Epidemiologic Study of the Elderly (H-EPESE) [11]; however non-significant findings for diabetes and CImp have been reported in the New Mexico Elder Health Study [19] in which cognitive function was also measured with the MMSE. It is thus possible that many of the earlier-reported research findings that link diabetes with CImp may actually be an artifact of other intervening phenomena such as regional and ethnic differentials in prevalence rates for diabetes. Thus, it is yet to be determined whether the relationship between diabetes and CImp is clinically relevant.
No such doubts exist about the relationship between cognitive and functional and impairments. Results reported in Table 1 (Blocks 3, Subtable 1A, Subtable 1B, and Subtable 1C) indicate that for each unit increase in ADLimp there is a corresponding 1.33 unit increase in odds for MMSE impairment. Corresponding ADL findings for CLOXI and CLOX2 are 1.22 and 1.21. The significant nature of these associations is compatible with those reported in a large number of other studies, including the above-cited. The only debate in this area appears to be causal direction. Is ADLimp a derivative of CImp, or is this relationship more reciprocal in nature? [17] Whatever the causal connection, these statistically significant findings are clinically relevant to medical professionals who deal with patients undergoing cognitive decline.
A growing body of research suggests that Hispanic ethnicity itself may constitute a risk factor in CImp [34-36]. Indeed, [37] suggest that Hispanics (like African Americans) may have a greater genetic risk of developing Alzheimer's disease than NHWs; and Mexican Americans may possess a relatively high contingent of risk factors associated with vascular cognitive impairment, including cerebral microvascular disease [22,38,39].
Our findings lend some degree of support to these assertions. After controlling for effects of all other variables in the equation (Blocks 3 of Subtable 1A, Subtable 1B, and Subtable 1C) El Paso's elderly Mexican Americans possess odds 2.46 times greater than NHWs for MMSE impairment, 1.53 times greater for impairment in executive function (CLOXI), and 2.35 times greater for CLOXII impairment. Again, these findings may be clinically relevant to medical professionals whose practices involve routine physical examinations for adult Mexican-American patients.
Finally, results in this study point to the importance of utilizing a number of different screening devices for assessment of cognitive function. The author of the CLOX drawing tests [40] suggests that CLOXI is more sensitive to executive function and to dementia presentations similar to those found in Alzheimer's disease than are other clock drawing tests. The CLOX drawing tests are also less sensitive than the MMSE to combined effects of education, SES, acculturation, language, and gender [27], and more able than the MMSE to discover the mild cognitive impairment syndromes found in non-Alzheimer's type dementias [41]. Thus, similar and significant results across multiple measures of cognitive function can discriminate among a variety of dementia-like presentations and increase the likelihood that these results can be taken as valid, dependable, and clinically meaningful.
This is the first study to evaluate the relationships between diabetes, Parkinsonism, and cognitive impairment in the context of Hispanic vs. Non-Hispanic ethnicity.
This study is supported by grants from the Paso del Norte Foundation's Center for Border Health Research, the United States Department of Health and Human Services Administration on Aging, and the Hogg Foundation for Mental Health.