Olfaction and Physical Functioning in Older Adults: A Longitudinal Study

Abstract

Background

Poor olfaction is associated with worse functional performance in older adults, but longitudinal evidence is lacking. We investigated poor olfaction in relation to longitudinal changes in physical functioning among community-dwelling older adults.

Method

The analysis included 2 319 participants from the Health, Aging and Body Composition study (aged 71–82 years, 47.9% men, and 37.3% Blacks) who completed the Brief Smell Identification Test in 1999–2000. Olfaction was defined as good (test score 11–12), moderate (9–10), or poor (0–8). Physical functioning was assessed up to 4 times over 8 years, using the Short Physical Performance Battery (SPPB) and the Health Aging and Body Composition Physical Performance Battery (HABCPPB). We conducted joint model analyses and reported the differences in annual declines across olfaction groups.

Results

During the follow-up, compared to those with good olfaction, older adults with poor olfaction had greater annual declines in both the SPPB score (−0.137, 95% CI: −0.186, −0.088) and all its subscales: standing balance (−0.068, 95% CI: −0.091, −0.044), chair stand (−0.046, 95% CI: −0.070, −0.022), and gait speed (−0.022, 95% CI: −0.042, −0.001). A similar observation was made for the HABCPPB score (difference in annual decline: −0.032, 95% CI: −0.042, −0.021). These findings are robust and cannot be explained by measured confounding from demographics, lifestyle factors, and chronic diseases or by potential biases due to death and loss of follow-up. Similar associations were observed across subgroups of sex, race, and self-reported general health status.

Conclusion

This study provides the first epidemiological evidence that poor olfaction predicts a faster decline in physical functioning. Future studies should investigate potential mechanisms.

Poor olfaction is a common sensory deficit in older adults, affecting about 15%–25% of adults aged 65 years and older (1–3). In older adults, poor olfaction has been best studied as one of the most important prodromal symptoms of Parkinson’s disease (PD) and dementia (4,5), and as a robust predictor for all-cause mortality (6–8). Interestingly, we recently reported that dementia and PD explained only a small portion of the excess mortality in older adults with poor olfaction (8), suggesting poor olfaction may have broad ramifications for the health of older adults, most of which are unknown.

Maintaining physical function is crucial to successful aging (9), whereas poor physical functioning predicts the loss of independence, increased morbidity, hospitalization, and mortality (10–13) in older adults. Although barely investigated (14–17), poor olfaction may lead to declines in physical functioning as people age. For example, poor olfaction may contribute to changes in diet and nutrition, which over time may lead to weight loss and poor physical functioning in the context of aging. However, to our knowledge, only a few cross-sectional studies have examined olfaction in relation to physical functioning in aging populations (14–17). We hereby conduct the first longitudinal study to assess the relationship of poor olfaction to changes in physical functioning over 8 years in a biracial, community-based cohort of older adults.

Method

Study Population

The Health, Aging and Body Composition (Health ABC) study was designed to investigate whether body composition changes act as a common pathway by which multiple diseases affect illness, disability, and mortality in older adults. Details of the study were published elsewhere (18). Briefly, 3 075 well-functioning, community-dwelling older adults (48.4% men and 41.6% Blacks) aged 70–79 years were enrolled from 1997 to 1998 in Pittsburgh, Pennsylvania and Memphis, Tennessee. Eligibility criteria included no reported difficulty walking a quarter mile, climbing 10 steps, performing daily living, no active cancer treatment in the previous 3 years, and no intention to move away from the study area for the next 3 years. Study participants were followed with annual clinic or home visits and semiannual or quarterly telephone interviews. The sense of smell was tested with the 12-item Brief Smell Identification Test (B-SIT; Sensonics, Haddon Heights, NJ) at the Year 3 clinical examination in 1999–2000. Physical functioning was assessed using the Short Physical Performance Battery (SPPB) and the Health ABC Physical Performance Battery (HABCPPB) at the Year 4, 6, 10, and 11 clinical visits.

In the present analysis, we defined the Year 3 clinical visit when B-SIT was taken (1999–2000) as the baseline, and Year 11 when the physical function was last assessed (2007–2008) as the end of follow-up. A total of 2 537 participants had valid olfaction test data at baseline. After excluding participants missing on covariates (n = 36) and lacking SPPB (n = 182) or HABCPPB (n = 183) measures at all clinical visits of years 4, 6, 10, and 11, the analysis of SPPB included 2 319 participants and the analysis of HABCPPB included 2 318 participants (Supplementary Figure 1). Compared with participants in the analyses, those excluded were slightly older, less likely to be Whites, from the Pittsburgh site, nonsmoker and physically active, and more likely to report lower than high school education, fair-to-poor health, dementia, cardiovascular diseases (CVDs), diabetes, chronic kidney disease, and depression (Supplementary Table 1). We also presented the number of participants with available SPPB and HABCPPB measures by olfaction groups at each clinical visit in Supplementary Table 2.

Olfaction Assessment

The B-SIT is a short version of the 40-item Pennsylvania Smell Identification Test and has been widely used to screen for olfaction impairment in clinical and epidemiologic studies (19–21). The test asks participants to scratch and smell 12 common odorants, one at a time, and identified the correct odorant from 4 possible answers in a forced-choice format (22). Each correct answer gets 1 point, and the total score ranges from 0 to 12, with a higher score indicating a better sense of smell. In this analysis, we defined olfaction as poor (B-SIT score 0–8), moderate (9,10), and good (11,12). This definition corresponds to the B-SIT score distribution tertiles and is consistent with population norms (19) and uses in other epidemiologic studies (20,23).

Physical Functioning

The SPPB is a well-established battery to assess the lower extremity physical function of older adults (13,24). It assesses the performance of standing balance, lower extremity strength, and gait speed, each rated on a scale of 0 (unable to do) to 4 (best performance) (24). The tests of standing balance include the side-by-side, semi-tandem, and full-tandem stands, each for 10 seconds. The lower extremity strength is assessed by completing 5 repeated stand-ups and sitting down from a chair without using arms for assistance. The gait speed is measured by the time to walk 6 m at a usual pace. The overall SPPB score ranges from 0 to 12, with a higher score indicating better performance. As the Health ABC participants were overall well-functioning at enrollment, the study investigators developed the HABCPPB scale to better assess the physical capacities of good performers in order to minimize the potential ceiling effects of using SPPB in the study population (18). The HABCPPB increased the holding time of semi- and full-tandem stands of SPPB to 30 seconds and added a 30-second single leg stand and a 6-m narrow walk of balance test (18). The HABCPPB score was scaled as a continuous variable, ranging from 0 to 4, with a higher score for better performance.

Baseline Covariates

We selected the following variables as potential confounders based on the literature (14,25): age, sex, race, study site, education, smoking, alcohol drinking, brisk walking as a surrogate for physical activity, body mass index (BMI), self-reported health status, and major chronic diseases. Covariates were from the Year 3 clinical visit, except for the study site, sex, race, education, and alcohol drinking obtained from Year 1. We defined chronic diseases based on previously published criteria. Briefly, CVDs included adjudicated diagnoses of coronary heart disease, congestive heart failure, cerebrovascular disease, and peripheral vascular disease (26). Diabetes was defined by a patient-reported diagnosis by a physician, antidiabetic medication use, a fasting blood glucose ≥6.99 mmol/L, or an oral glucose tolerance test result ≥11.10 mmol/L(27). Chronic kidney disease was defined as a glomerular filtration rate <60 mL/min/1.73 m² (28). Lung diseases, including pneumonia, asthma, bronchitis, chronic obstructive pulmonary disease, and emphysema, were adjudicated mainly based on hospitalization data, assisted by reviews of available x-ray, spirometry tests, and physical exams (29). Depressive symptoms were defined as a score ≥10 on the 10-item Center for Epidemiologic Studies Depression Short Form (30). Dementia was defined as a Modified Mini-Mental State examination score <80 at baseline, a race-stratified decline of ≥1.5 SDs afterward, or an adjudicated diagnosis of dementia based on hospitalization and medication data (5). PD cases were ascertained via expert adjudication using data of self-reported PD diagnosis by a physician, medication uses, and adjudicated causes of hospitalization and/or death (4).

Statistical Analysis

In all analyses, we used participants with a good sense of smell as the reference group. We used chi-squared and analysis of variance tests to compare baseline characteristics across olfaction status. We conducted joint model analyses (31) to assess the association of olfaction with changes in physical functioning. By combining a survival submodel with a longitudinal submodel, this joint modeling approach enabled us to account for potential biases from informative missing due to death or loss of follow-up (32). Specifically, we fitted a Cox-type cause-specific hazard model for the competing risks of death or dropout. In this survival submodel, death and dropout were modeled separately. We then fitted a linear mixed-effect submodel for each physical functioning measure, with subject-specific random intercept and slope. Participants were followed from baseline until the date of death, last eligible visit, or the end of the follow-up, whichever came first. We tested the linearity assumption by adding a quadratic time variable in the fixed-effects part of the mixed models, and it was not statistically significant in either the SPPB or the HABCPPB model. We, therefore, assumed a linear trajectory of physical functioning in the analyses.

We first assessed the marginal or crude association of olfaction and physical functioning, and then sequentially adjusted for 3 sets of potential confounders. Model 1 only adjusted for age, sex, race, and study site; model 2 added education, smoking, alcohol drinking, brisk walking, and BMI; and model 3 further added general health status, CVDs, diabetes, chronic kidney disease, lung diseases, depressive syndromes, and dementia or PD. We also conducted subgroup analyses by sex, race, and self-reported general health status. Our prior analyses in the Health ABC study showed that men and Blacks were about twice more likely to have poor olfaction than women and Whites (3), the associations of poor olfaction with PD (4) and dementia (5) were likely to be stronger in men and Whites than in women and Blacks, and the association of poor olfaction with excess mortality was largely limited to participants who had reported good-to-excellent health (8). Finally, to further examine whether the association was confounded by neurodegenerative diseases at baseline, we conducted sensitivity analyses among participants with no dementia or PD at baseline.

In all longitudinal submodels, we included an interaction term between olfaction and time. The coefficients for olfaction quantified the between-group differences of physical functioning at baseline, while the coefficients for the interaction terms represented the between-group differences of annual change in physical functioning. All survival submodels used the same sets of covariates as those in the longitudinal submodels. Because this analysis focuses on the trajectories of physical functioning in relation to olfaction, we did not provide the survival submodel results. Besides, we provide both estimates from the joint model and the linear mixed-effect submodel to show the effects of inclusion of the survival submodel for our main results. All statistical analyses were performed with the SAS statistical software, version 9.3 (SAS Institute, Inc., Cary, NC) and the R package JM (33) version 1.4-5 with a 2-sided test and α of .05.

Results

The mean age of study participants was 75.6 ± 2.9 years (range: 71–82) when their sense of smell was tested. Compared to participants with good olfaction, those with poor olfaction were older and more likely to be men, Black, from the Memphis study site, and ever smokers. They were also more likely to report lower education, lower BMI, and fair-to-poor overall health. For the chronic diseases examined, poor olfaction was associated with prevalent dementia, PD, chronic kidney disease, and depressive symptoms (Table 1).

Olfaction and SPPB

Compared with older adults with good olfaction, those with poor olfaction tended to have a lower SPPB score expected at baseline when their olfaction was tested (Table 2). The difference was −0.433 (95% CI: −0.677, −0.189) after adjusting for demographics, but substantially attenuated to −0.161 (95% CI: −0.396, 0.075) and lost statistical significance after further adjusting for lifestyle factors and chronic diseases. Similar observations were also made for the difference between moderate and good olfaction.

During the follow-up, the observed SPPB scores declined over time across all olfaction groups (Figure 1A). Participants with poor olfaction had a greater decline than those with good olfaction (Table 2). The between-group difference in decline is −0.141 (95% CI: −0.190, −0.091) per year after accounting for demographics, and the effect estimate barely changed with additional adjustment for lifestyle and health-related conditions (−0.137, 95% CI: −0.186, −0.088). The SPPB of participants with moderate olfaction declined slightly faster than that of the reference group, but the difference did not reach statistical significance (−0.044, 95% CI: −0.089, 0.002) in the fully adjusted model (Table 2). Estimates from the linear mixed-effect submodel (Supplementary Table 3) were similar to those from the joint model. Similar associations were found in subgroup analyses by sex, race, and general health status (Supplementary Figure 2). In the sensitivity analysis, the association was only modestly attenuated after excluding participants with baseline dementia or PD (Supplementary Table 4). For example, the difference in annual decline was −0.120 (95% CI: −0.172, −0.067) between the poor versus good olfaction group in the fully adjusted model.

Olfaction and SPPB Subscale

Of the 3 SPPB subscale measures expected at baseline, only gait speed showed significant associations with olfaction: compared to participants with good olfaction, the fully adjusted score was −0.169 points (95% CI: −0.277, −0.062) lower for those with poor olfaction, and −0.127 (−0.228, −0.026) lower in moderate olfaction at baseline (Table 3).

As to the changes over time, all subscale scores decreased faster in participants with poor olfaction than in the reference group. In the fully adjusted model, annual between-group differences in decline were −0.068 (95% CI: −0.091, −0.044) for standing balance, −0.046 (95% CI: −0.070, −0.022) for chair stand, and −0.022 (95% CI: −0.042, −0.001) for gait speed. Moderate olfaction was also associated with greater declines in standing balance (−0.029, 95% CI: −0.050, −0.007) and chair stands (−0.021, 95% CI: −0.043, −0.001). The analysis excluding dementia and PD showed similar results except that the longitudinal between-group difference in gait speed did not reach statistical significance (Supplementary Table 4).

Olfaction and HABCPPB

Overall, the analyses using HABCPPB as the outcome generated similar results to those that analyzed SPPB (Table 2; Supplementary Figure 3 and Supplementary Table 4).

Ethical Approval

The Health ABC study was approved by the institutional review boards of the 2 field centers (University of Pittsburgh and University of Tennessee, Memphis) and that of the coordinating center, the University of California, San Francisco. All participants signed an informed written consent, approved by the institutional review boards at the clinical sites. This specific analysis was exempted by the institutional review board at the Michigan State University.

Discussion

In this large biracial cohort of community-dwelling older adults, poor olfaction was associated with a greater decline in physical functioning over 8 years. To the best of our knowledge, this study provides the first epidemiological evidence that poor olfaction predicts a faster decline in physical functioning. This observed association was independent of potential confounding from population demographics, lifestyle factors, and major health indicators; neither can it be explained by potential bias due to death and loss of follow-up. This association is found in both men and women, Blacks and Whites, and people with different health statuses, suggesting robustness and generalizability.

About 15%–25% of older adults in the United States (~7.4–12.3 million) have a poor sense of smell (1–3). Unlike vision or hearing loss, two-thirds of the older people with poor olfaction are unaware that they have this sensory defect (1–3). The impacts of poor olfaction on aging and the health of older adults are yet to be fully defined. To date, it has been best studied as one of the earliest and most important prodromal symptoms of dementia and PD (3,4,34). Interestingly, we found that dementia and PD only accounted for 22% of the 10-year excess mortality associated with poor olfaction in older adults (8). Coupled with another 6% explained by weight loss, about 70% of this higher mortality remains unexplained (8). These findings lead us to explore the broader health ramifications of poor olfaction beyond neurodegenerative diseases.

As the sense of smell, physical functioning also declines with age. In older adults, physical functioning is central to maintaining their health and quality of life (35,36). Many studies have linked poor physical functioning to the loss of independence, hospitalization, as well as increased morbidity and mortality (10–13). In this study, we assessed the association of olfaction with physical functioning using 2 objective measures. The SPPB is a commonly used, standard and clinically relevant index of lower extremity physical functioning in older adults (37,38), covering 3 out of 5 National Institutes of Health Toolbox domains for motor assessment (locomotion, balance, and strength) (39). It is sensitive to changes in physical performance over time (38,40,41), and strongly predicts adverse health outcomes of older adults (24,42,43). The HABCPPB was developed to minimize the potential ceiling effects of the SPPB and has been widely used in analyzing the Health ABC data. Consistent findings using 2 solid outcome measures support the validity of our results.

To our knowledge, the association of poor olfaction with physical functioning has been only examined in several cross-sectional and relatively small studies. In a cross-sectional analysis of 163 participants from the Baltimore Longitudinal Study of Aging, poor olfaction measured with the 16-item Sniffin’ Sticks was associated with poor motor functions, including mobility, balance, and fine motor function (14). This finding was supported by 2 other cross-sectional analyses (n = 768 and 1 035, respectively), which reported self-reported olfaction was associated with frailty in older adults (15,16). Our study is much larger than previous studies and assessed physical functions longitudinally with objective measures. Further, we conducted comprehensive analyses to control for confounding and biases and examined potential heterogeneity via subgroup analyses. We identified a robust association of poor olfaction with declines in physical functioning in older adults that cannot be explained by a range of measured confounding and biases. Although we cannot directly establish the clinical significance of our observation, a previous Health ABC analysis suggests clinical relevance of a decline in the SPPB score of 0.31–0.78 points over 3 years (38).

Therefore, evidence to date, albeit preliminary and limited, consistently supports an association of poor olfaction with poor physical functioning. Although the underlying mechanisms are yet to be ascertained, there are several plausible explanations. First, we reason that poor olfaction may lead to declines in physical functioning by reducing appetite, promoting poor food choices (44), and resulting in faster loss of lean mass (45). Second, our finding may in part be explained by preclinical declines in the physical functioning of older adults in prodromal stages of neurodegenerative diseases. Poor olfaction has been linked to both AD and PD pathologies in people without a clinical diagnosis (46–48). Further, in older adults with no clinical neurodegeneration, poor olfaction is associated with structural abnormalities in brain regions related to spatial navigation, sensorimotor regulation, mobility, and balance (14,49). Finally, both olfaction and physical functioning decline with age, and therefore, the observed association may be partly due to the underlying biology of aging. Although our study participants have a relatively narrow range of chronological age and we carefully adjusted for it in the analysis, we cannot exclude the intriguing possibility that both poor olfaction and physical functional decline are the results of accelerated aging beyond its chronological pace. These potential explanations for the observed association are not mutually exclusive and are subject to further mechanistic investigations.

Our study has several notable strengths. The Health ABC study is a well-established, community-based biracial cohort that was designed to study functional declines in older adults. It has one of the best collections of physical functioning data, repeatedly assessed over the years, enabling both cross-sectional and longitudinal investigations. The sample size is relatively large, allowing us to conduct comprehensive statistical analyses to examine both the validity and generalizability of our findings (eg, analyses by sex and race). Our study also has several limitations. First, the study participants were fairly old at the time of olfaction testing, which may limit the generalizability to relatively young older adults. Second, compared to participants in the analyses, those who were excluded were slightly older and generally less healthy. Although we have conducted comprehensive multivariate and stratified analyses, for example, by self-reported health status, we could not exclude the possibility of bias from this source. Third, the sense of smell decreases with age, but we assessed olfaction only once at a single time point. Future studies might be interested in examining the trajectory of olfaction decline in older adults and how it correlates with physical functioning and health outcomes. Fourth, this B-SIT test itself does not differentiate chronic from the acute onset of poor olfaction nor its causes. However, given the age of our study participants, we suspect a vast majority of the poor olfaction in our study is likely aging-related and chronic. Finally, despite our efforts to control for a wide range of potential confounders, our analyses are subject to unmeasured and residual confounding, as in all observational studies.

In conclusion, we found that poor olfaction in older adults was associated with a greater decline in physical functioning. Future studies should investigate the underlining mechanisms, their health implications, and preventive strategies in older adults.

Funding

This study was supported by a start-up fund from the Michigan State University (GE100455). The Health ABC study was supported by the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR), the Intramural Research Program of the NIA/National Institutes of Health and NIA contracts N01AG62101, N01AG62103, and N01AG62106, NIA grant R01AG028050, and NINR grant R01NR012459.`

Conflict of Interest

The authors have no conflicts of interest to disclose. All authors have declared no financial relationships with any organizations that might have an interest in the submitted work in the previous three years and no other relationships or activities that could appear to have influenced the submitted work.

Author Contributions

H.C. conceived the study. E.J.S. and E.M.S. provided data. Y.Y., Z.L., C.L., and H.C. designed and performed statistical analyses. Y.Y. and H.C. prepared the first draft of the manuscript and all other coauthors provided critical comments. All authors have contributed to the interpretation of results, critical revision of the manuscript, and the manuscript prepration and gave final approval for submission.

References