Informant-reported cognitive symptoms that predict amnestic mild cognitive impairment

The process of differentiating age-associated memory decline from those who might have a clinically significant disorder of memory and cognition is difficult. In particular, distinguishing individuals with amnestic mild cognitive impairment (aMCI) from those who are cognitively normal (CN) is challenging, as memory and cognitive complaints are often reported in both groups from both the patient and informants [1]. Given that the current diagnostic criteria for aMCI include subjective (patient and/or family report of decline) and objective (neuropsychological testing) evidence of memory decline, a clinician's initial impression from a relatively short office visit may not allow for an accurate assessment [2].

Amnestic MCI was first characterized as a syndrome consisting of memory performance at or below 1.5 standard deviations (SD) on age- and education-adjusted normative values on a verbal memory test along with subjective memory complaints, preserved global cognition, and preserved activities of daily living [3]. The diagnostic criteria for MCI have since been refined to differentiate between amnestic and non-amnestic forms, with the latter showing performance at or below 1.5 SD on a test or test(s) in one or more domains other than memory. Both amnestic and non-amnestic MCI can be further classified as single or multiple domain MCI depending upon the number of cognitive domains that show test performance(s) at or below 1.5 SD [4].

Several studies have investigated the clinical course and presentation of individuals who have self- and informant-reported memory complaints [5‐8]. Some evidence suggests that individuals who are cognitively normal and have subjective memory complaints demonstrate MRI findings that are similar to those of aMCI individuals [9]. Other studies have demonstrated that an informant's report of an individual's cognitive status is valid and highly accurate in the very early stages of AD [6]. Although the diagnostic criteria for aMCI do not include functional impairment, previous studies have found that aMCI patients may have difficulty with higher level daily activities, such as balancing a checkbook, and may show mild, but not significant, difficulty in daily functioning [1, 10].

Utilizing additional information with added discriminatory power can aid in identifying individuals at risk for developing Alzheimer's disease (AD), a task of greater interest now, with emerging early AD treatments [10, 11]. To accomplish this, identifying certain cognitive symptoms that may yield greater diagnostic accuracy than subjective memory complaints alone is necessary. A recent pilot study found that the Alzheimer's Questionnaire (AQ), an informant-based questionnaire designed for use in primary care settings, has both high sensitivity [87.00 (77.00 - 94.00)] and specificity [94.00 (84.00, 99.00)] for aMCI [12].

The intent of this study is to determine which AQ items are predictive of aMCI. By identifying cognitive symptoms beyond subjective memory complaints, individuals at risk for developing AD may be identified more quickly so that further diagnostic testing and subsequent treatment may be initiated sooner in the disease process.

Demographic characteristics of the study sample are displayed in Table 1. The CN group was older and slightly more educated than the aMCI group. Males and females had relatively equal representation across groups. Chi-square analysis showed significant differences in response frequencies for all but two AQ items (Table 2). Results from the multiple forward stepwise logistic regression analysis, which adjusted for age and education, are displayed in Table 3; only the AQ items that were included in the stepwise model are shown. This model yielded four AQ items as strong predictors of aMCI, which are listed in Table 3 with their associated odds ratios (OR), 95% confidence intervals, and p-values. The resulting stepwise logistic model yielded a Nagelkerke R² of 0.71 indicating that a large proportion of the variance between aMCI and CN individuals was accounted for by the four AQ items.

In order to more accurately characterize the clinical validity of these findings, a second non-stepwise logistic regression analysis was carried out which used only the four significant AQ items while correcting for age and education. This model yielded sensitivity of 80.30 (67.00, 89.53) and specificity of 81.80 (69.67, 90.37) with an area under the curve (AUC) value of 0.94 (0.89, 0.99).

The results of the study indicate that the four informant-reported items listed immediately above are highly predictive of aMCI. These items are memory-related, and also suggest some degree of impairment in higher-level functional abilities. The use of informant-supplied information is a widely-used and highly valid method of assessing an individual's cognitive and functional abilities [5, 7]. Relative to other informant-based instruments [19‐22] the AQ takes substantially less time to administer [12], a fact of importance to clinicians with very limited time [23].

For clinicians who see patients with subjective memory complaints, accurate identification of those who need further evaluation is critical to cost containment and resource management. A significant proportion of older adults present with subjective memory complaints [24, 25], and these complaints can precede the onset of clinical AD [26]. The large and growing number of older adults underscores the importance of utilizing brief and accurate screening measures. Additionally, as new therapies for AD transition from being symptomatic to disease-modifying, identifying individuals who are at-risk or are in the earliest stages of the disease will be crucial in determining and improving disease outcome [1].

There are some limitations to this study. The first is that the confidence intervals for the odds ratios of the statistically significant AQ items were relatively wide, indicating decreased statistical power. Although the sample was large enough to yield robust odds ratios for the four AQ items, a larger sample size might provide a more accurate estimate of effect size. In addition, the R² value may not truly represent the amount of variance accounted for by the model. The reason for this is that R² values in logistic models are approximations of linear-based R² measures and are not fully equivalent. In addition, R² measures used in logistic models are prone to bias when used with small sample sizes and may result in an inflated estimate of the amount of variance accounted for [27]. Another limitation is that the AQ itself requires the use of an informant. In some cases, a patient may come to a physician's office alone or they may not have a reliable informant available to do the assessment. Although several patient-based cognitive assessments, such as the Mini Mental State Exam [28], can be used, they are subject to confounding factors such as cultural effects and low education [29‐31]. Finally, the study sample was homogenous with respect to ethnicity, as all subjects were Caucasian, so it is unclear whether these results are applicable to an ethnically diverse population.

In addition, the ability of other widely-used informant-based instruments to accurately identify clinical aMCI has not been established. The validity and accuracy of the AD8 has been established in clinical AD and in individuals with a CDR global rating 0.5 which is considered "very mild dementia" [32]. It is important to note that this categorization (CDR = 0.5) does not necessarily equate to a clinical diagnosis of aMCI so it is uncertain whether the AD8 can accurately identify clinically-defined aMCI cases. In addition, a recent study demonstrated that the IQCODE does not have high sensitivity in the detection of aMCI [33]. As mentioned earlier, a previous pilot study of the AQ demonstrated high sensitivity and specificity for aMCI when compared to cognitively normal individuals. The results of the current study showed that four statistically significant AQ items accounted for large proportion of the variance between aMCI and CN individuals and also yielded high sensitivity and specificity in differentiating the two groups. Overall, the results of this study indicate that certain AQ items can differentiate individuals with aMCI from those experiencing age-associated changes in memory and cognition. As assessed by the AQ, difficulties with orientation to time, repetition of questions and statements, difficulties in managing finances, and visuospatial disorientation were all significant predictors of aMCI as diagnosed by an expert in memory disorders.

Given that memory complaints are commonly reported by elderly patients and their family members [7], a means to quickly and accurately identify individuals who may be in the early stages of AD and in need of further evaluation is critical to not only cost containment and resource management, but also to earlier diagnosis in order to improve disease outcome. These data indicate that problems with orientation to time, repeating statements and questions, difficulty managing finances, and trouble with visusospatial orientation may accompany memory deficits in aMCI. From a clinical standpoint, these findings are important as it will allow clinicians to more easily and accurately determine which individuals require further assessment of cognitive problems.