Associations Between Plasma,Imaging,and Cerebrospinal Fluid Biomarkers with Driving Behavior and Cognitive Tests: Implications for Biomarker Usefulness

INTRODUCTION

The ATN research framework of the National Institute on Aging – Alzheimer’s Association (NIA-AA) depends upon “positivity” of markers of amyloid, tau, and neurodegeneration to operationalize the presence of biological Alzheimer’s disease (AD), even among persons with normal cognition [1]. Cut-offs to determine biomarker positivity were originally developed by finding the numerical point that best distinguishes a group with clinical dementia from one with normal cognition [2]. Validation and cut-offs for newer biomarkers, such as those derived from blood plasma, have also relied on their strength of association with older molecular biomarkers such as amyloid imaging [3, 4]. In addition to being able to distinguish clinical groups, these established cut-offs can successfully predict time to incident dementia among persons who are cognitively normal[5, 6].

The predictive utility of AD biomarkers, and their associated cut-off values, likely depends on what is being predicted. AD impacts widespread functional activities of daily living, such as driving [7, 8], which incorporates cognitive functioning. Driving is important for maintaining mobility and independence among older adults, and 83% of those aged 70 and above have current driver’s licenses [9–11]. However, traffic crashes are among the leading causes of accidental disability and death in this age group, with almost 700 persons aged 65+ years being injured and over 20 being killed every day in the US [12]. Being able to identify who is likely to be an unsafe driver, and who is at risk for future driving problems, will enable early intervention and reduce disability and death among olderadults.

The relationship between AD biomarkers and cognition differs based on both the specific biomarker used and the domain of cognition examined [13, 14]. Although driving behavior has been shown to be associated with traditional imaging and CSF biomarkers [8, 16], it is possible that, as with cognition, some biomarkers have stronger associations with particular aspects of driving behavior than others. Further, associations between newer plasma biomarkers and driving behaviors are currently unknown. Blood-based biomarkers are expected to play an important clinical role in the future, as they are likely to be more accessible, affordable, and acceptable to patients compared to biomarkers derived from CSF and imaging [17].

As a necessary first step in optimizing identification and prediction of AD-related driving difficulties, we examined the extent to which individual plasma, imaging, and CSF biomarkers are related to specific driving behaviors among cognitively normal older drivers. We also compared these associations to those of the biomarkers with selected cognitivetests.

MATERIALS AND METHODS

RESULTS

There were 167 participants that met the inclusion criteria for analyses. Each participant received a CDR = 0 at their clinical assessment closest to data logger installation (mean±standard deviation [SD] = 156.1±125.5 days). Participants’ ages ranged from 65.6 to 90.8 with a mean±SD of 73.3±4.9 years, a mean±SD of 16.6±2.2 years of education, 47.3% were women, and 13.8% were Black or African American (all other participants were White). Driving time with the data logger installed ranged from 0.6 to 4.7 (mean±SD = 2.2±0.9) years.

All participants had complete driving, clinical, and demographic data available, but the number with each biomarker type varied as follows: N = 139 with amyloid imaging; 103 with tau imaging; 144 with normalized hippocampal volume; 125 with CSF Aβ₄₂, Aβ₄₀, tau, and ptau₁₈₁; 118 with CSF NfL; and 116 with plasma Aβ₄₂, Aβ₄₀, and NfL. Table 1 shows the magnitude and direction of the Pearson correlation coefficients with each of the driving and cognitive variables. When analyses adjusted for age, gender, race, and education, the relative strength and direction of associations was similar, although fewer pairings were significant at the 0.05 or 0.01 alpha level (Table 2). This was expected, given the decrease in statistical power with greater numbers of independent variables in the models. Unless otherwise specified, the findings reported going forward refer to the adjusted analyses (Table 2).

DISCUSSION

The current study is a first attempt at understanding the interconnectedness between traditional and novel AD biomarkers, driving behaviors and cognition. A key finding of the study is that plasma biomarkers, like CSF and imaging biomarkers, are useful in predicting everyday behaviors on complex functional tasks such as driving. More specifically, these results show that elevated levels of plasma Aβ₄₀ and/or Aβ₄₂ are associated with fewer total trips and night trips, decreased trip distance and duration, as well as more speeding incidents. Taking a closer look at this interconnectivity, our findings further suggest that the associations between plasma biomarkers and driving behaviors differ from the relationship between these biomarkers and cognition. In fact, it can be observed that plasma Aβ₄₂/Aβ₄₀, though an effective predictor of cognition [21], may not be strongly associated with driving behaviors.

Another key finding of this study is that different AD biomarkers are associated with different dimensions of driving and cognitive functioning. It has long been shown that different AD biomarkers have varying associations with cognitive domains and specific domains (e.g., memory, attention, and executive function) [14]. Our results align with the broader observation that there is not always a strong association between cognitive performance and fluid markers of preclinical AD. Notably, this is consistent with observations from our earlier investigations [16, 27]. As the understanding of these dynamics expands, it could provide valuable insights into the nuanced progression of Alzheimer’s disease and its broader clinical manifestations. On the other hand, our findings provide evidence that different biomarkers are related to different everyday driving behaviors. For instance, elevated levels of CSF tau and NfL (reflecting neurodegeneration) are associated with more self-regulating driving behaviors. More specifically, higher levels of CSF tau were related to shorter duration of trips, while higher CSF NfL values were associated with traveling shorter distances, smaller radius of gyration and fewer trips at night. Furthermore, CSF and plasma Aβ₄₂ and Aβ₄₀ behave differently with respect to driving within the same individuals and may therefore yield complementary information. Overall, these distinct relationships may help in understanding how different biological changes that occur during the preclinical stage of AD can impact various sensorimotor and cognitive processes.

Our findings have implications for models of AD development, as they may help in establishing the timeline of the appearance of functional changes relative to changes in biomarkers and cognition. Additionally, biomarker cut-offs have been established because they are best at discriminating persons with symptomatic AD from those with normal cognition. However, these results indicate that the optimal cut-off points for biomarker positivity and negativity must be tailored to the specific outcome being monitored and predicted. Similar to impairment in cognitive processes, functional impairment is also a defining characteristic of AD as the disease progresses. Future AD research should work to determine the most effective cut-off values for using biomarkers to identify and predict functional outcomes such as driving, in order to provide a more nuanced understanding of the AD’s impact on daily living.

Though more work lies ahead before we can monitor, detect, or predict AD in the preclinical phase on an individual basis, the findings of this study open up other interesting directions for further research in the field. One future research direction is to investigate how AD pathologies, as reflected in different biomarkers, affect other daily activities such as falls and gait patterns. With appropriate techniques and devices, we can create a novel scheme for behavioral staging of AD across its entirespectrum.

Finally, the findings should be considered in light of a few limitations. First, testing these relationships in additional samples are required to establish the reliability of these findings. Second, the sample lacks generalizability since the participants in this study are almost all non-Hispanic White and primarily reside in the St. Louis Area. Similar studies should be conducted in more geographically and racially diverse samples to establish the generalizability of these findings. Additionally, although our small sample size was effective for demonstrating that individual AD biomarkers (plasma, CSF, and imaging) are differentially related to naturalistic driving behaviors and cognitive test scores, the application of larger sample sizes, with increased statistical power, might uncover associations that were not deemed significant in our final models. For example, CSF NfL was significantly correlated with acceleration in the unadjusted analyses, but not in the GLMs.