With Teeth,Broken,or Fixed: The Challenges of Linking Periodontitis,Neuroepidemiology,and Biomarkers of Disease

Periodontitis, an inflammatory disease of the tooth-supporting structures elicited by dental plaque biofilms [1, 2], is a major cause of tooth loss in adults [3]. The disease is highly prevalent in older people [3–6], disproportionately affects disadvantaged populations [6, 7], results in frequent transient bacteremias [8, 9], and elicits systemic responses including circulating antibodies to periodontal microbiota [10–12] and inflammatory markers [13, 14]. Caries and periodontitis are the predominant causes of tooth loss leading to edentulism and masticatory dysfunction.

Several established risk factors for Alzheimer’s disease (AD) and related dementias (ADRD) include diabetes, obesity, and smoking, and induce shared systemic inflammatory responses also seen with periodontitis [15–18]. Chronic inflammation may have direct cerebral effects including AD expression [19, 20] and indirect effects through pathways involving accelerated atherogenesis [15, 21–23]. Previous epidemiologic studies have demonstrated that periodontitis and tooth loss predict prevalent cognitive impairment [24, 25], incident dementia [26–29], and dementia-associated mortality [30].

Many questions remain regarding the temporality and potential mechanisms through which periodontitis may potentially be part of a causal pathway for ADRD. Aside from chronic inflammation, additionally proposed mechanisms [31] include tooth loss leading to alterations in diet, and fewer inter-occlusal contacts decreasing somatosensory feedback to the brain. It also remains possible that periodontal-systemic associations may reflect residual confounding of complex socioeconomic factors [32].

A central concern to the field is that of reverse causality. It is thus unclear whether neurological problems are due to chronic, pre-existing oral health conditions, or whether the earliest changes of cognitive decline result in inattention to oral care leading to declining oral health [32, 33]. When exploring late-life or long-latency illnesses such as ADRD, life-course exposures to chronic conditions such as periodontitis require sufficiently large, longitudinal cohorts, with standardized assessments of both exposures and outcomes of interest. Few such studies are available to date.

In this issue of the Journal of Alzheimer’s Disease, Zhang et al. [34] explore the impact of various self-reported oral health conditions on incident dementia and MRI findings. Likely the largest periodontitis-neuroepidemiological study in ADRD to date, it leveraged participant registry data from the UK Biobank, and supports findings from earlier studies that have linked baseline metrics of oral health and incident cognitive decline. The study goes a step further than most by exploring radiographic outcomes of brain atrophy regionally implicated in ADRD.

This is not the first sizeable epidemiological study to explore associations between periodontal disease and both incident cognitive and MRI outcomes. Previously, the Atherosclerosis Risk in Communities (ARIC) cohort study analyzed baseline oral health-related variables in participants in their seventh decade of life and cognitive and MRI outcomes assessed at follow-up. ARIC, a four-site longitudinal US cohort, used standardized cognitive and comprehensive oral health assessments and identified associations between periodontitis and edentulism and incident cognitive impairment [35]. However, no associations were found in ARIC with MRI brain volumes, risk for infarcts, or amyloid-PET outcomes. Neither ARIC nor the UK Biobank study analyzed white matter hyperintensity volumes, an important MRI metric associated with ADRD. Importantly, the ARIC cohort had substantial loss to follow-up by the time of imaging, with approximately 22% completing MRI measures and just 4% having amyloid-PET imaging, with most imaging occurring about 15 years after the oral health examination. Appropriately, there were attempts to address issues resulting from missing MRI data and loss to follow-up due to death or censoring.

The UK Biobank is a large population-based study with linkage to national health care records. It is subject to a healthy volunteer bias relative to the general population but with study outcomes generally thought to remain valid among its participants [36]. The UK Biobank initially enrolled its participants from 2006 to 2010, and additionally began collecting standardized MRI assessments in 2014. In the present study, 38675 of 425183 of the cohort (9.1%) had MRI imaging available at the time of the analyses, highlighting issues again of missing data potentially impacting interpretation.

Despite the strengths and consistency with cognitive outcomes in other large epidemiological studies [37, 38], the study highlights and partly addresses problems and limitations common to other periodontitis-neuroepidemiological cohort studies: 1) measures of oral health may be added on to an established cohort but are available in a smaller subgroup, 2) neurological biomarker data are available in a subgroup, and 3) oral health metrics may be incomplete, subject to bias, or require adjustments [39, 40]. Further complicating matters, diagnostic criteria for periodontitis have themselves evolved over time [41, 42].

Finally, self-reported metrics of oral health are problematic and can reflect a mixture of awareness, access to care, or willingness to seek care when a problem is identified. In this study, available UK Biobank metrics were primarily limited to self-report of a handful of oral conditions, with some linkable to healthcare records. Upon self-assessment, some findings may be more obvious and objectively confirmable (ulcers, bleeding gums, and dentures) but validation relies on self-perceptions of oral health and ultimately seeking care. Nonetheless, there is some evidence of internal validation: self-reported measures were associated with caries and pulp and apical tissue diseases, and gingivitis and periodontal diseases were associated with multiple measures of self-report. Of the measures in the study, only bleeding gums and loose teeth are somewhat specific to current periodontal disease, while full-denture use reflects an end-stage condition.

The most meaningful inferences drawn from epidemiological studies of any kind arise when there are a priori plans to comprehensively record exposures and outcomes of interest. The challenge posed to studies such as the one highlighted here are that clinico-pathobiological underpinnings of both periodontitis and ADRD have only recently come into focus, on a parallel timeline with a growing scientific interest in possible mechanistic relationships. To date, it is established that there are complex relationships between common oral health conditions [6], the periodontal microbiome [43], host-responses [44, 45], and genetic factors [46] which together inform ongoing studies with available multimodal clinical and biomarker-based assessments of cognitive aging outcomes.