The Adaptive Olfactory Measure of Threshold (ArOMa-T): a rapid test of olfactory function

Abstract

Many widely used psychophysical olfactory tests have limitations that can create barriers to adoption. For example, tests that measure the ability to identify odors may confound sensory performance with memory recall, verbal ability, and prior experience with the odor. Conversely, classic threshold-based tests avoid these issues, but are labor intensive. Additionally, many commercially available tests are slow and may require a trained administrator, making them impractical for use in situations where time is at a premium or self-administration is required. We tested the performance of the Adaptive Olfactory Measure of Threshold (ArOMa-T)-a novel odor detection threshold test that employs an adaptive Bayesian algorithm paired with a disposable odorant delivery card-in a non-clinical sample of individuals (n = 534) at the 2021 Twins Day Festival in Twinsburg, OH. Participants successfully completed the test in under 3 min with a false alarm rate of 7.5% and a test-retest reliability of 0.61. Odor detection thresholds differed by sex (~3.2-fold lower for females) and age (~8.7-fold lower for the youngest versus the oldest age group), consistent with prior studies. In an exploratory analysis, we failed to observe evidence of detection threshold differences between participants who reported a history of COVID-19 and matched controls who did not. We also found evidence for broad-sense heritability of odor detection thresholds. Together, this study suggests the ArOMa-T can determine odor detection thresholds. Additional validation studies are needed to confirm the value of ArOMa-T in clinical or field settings where rapid and portable assessment of olfactory function is needed.

Keywords: COVID-19; adaptive algorithm; anosmia; detection threshold; sex differences; smell.

Fig. 1.

Flow diagram summarizing data cleaning steps resulting in the final participant set used for these analyses (n = 534).

Fig. 2.

The ArOMa-T. The test includes a bi-fold card graphics and text on the outside pages (A) and with user instruction and 17 peel-and-burst panels that contain varying amounts of the rose-like/floral odorant phenylethyl alcohol on the inside pages (B).

Fig. 3.

Detection threshold estimation in a single participant using ArOMa-T. (A) Schematic of the numbered peel-and-burst labels on the test card, shaded with gray to indicate differing PEA concentration from lowest (light) to highest (dark); on the actual card, users receive no cues about specific odorant concentration (see Fig. 1). Participants are asked whether they can smell the odor under label 1 (an intermediate concentration). Given a “Yes” response (cyan outline), the app directs them to label 14; a “No” response to this same label (not shown) would send them to panel 6 (a higher concentration that is easier to detect). A “No” response (magenta outline) for label 14 directs the participant to label 11, and so on. The sequence 1, 14, and 11 correspond to 1 possible path through the first 3 questions. (B) Based on psychometric theory and the participant’s responses (outlined in cyan and magenta for “Yes” and “No” responses, respectively, and corresponding to the labels outlined in panel (A)), the algorithm fits a psychometric curve that estimates the probability of a “Yes” response at all concentrations. The solid black line is the point estimate for that curve, and the shaded yellow region is the uncertainty (standard error). The odorant concentration at a “Yes” response probability halfway between the estimated minimum and maximum of that probability is identified as the threshold (vertical dashed line); the uncertainty in this value is indicated with the purple arrow. (C) The same participant completes an additional 5 trials in the same test (for a total of 8 trials); responses are again shown above and below the curve using colored rectangles (cyan square for “Yes,” magenta for “No”; these choices are not indicated in panel (A), as they could take a number of paths). With additional responses, the estimated curve (and threshold) has shifted slightly, and the uncertainty has been substantially reduced. For color figure refer to online version.

Fig. 4.

Smell thresholds using the ArOMa-T. (A) Raincloud plots showing odor detection thresholds stratified by sex. Open circles represent odor detection thresholds for individual participants; women are shown in purple and men in blue. (B) Raincloud plots showing odor detection thresholds stratified by age group, with each age group represented by a different color; again, open circles indicate odor detection thresholds for individual participants. In both panels, the solid black dot is the point estimate of the mean and error bars are 95% confidence intervals of that estimate. Group sample sizes are shown in parentheses on the left side of the plot. Thresholds shown in the boxes on the left and right sides were imputed in accordance with the ASTM E679 rules for extreme values outside the range of the concentrations tested; the light blue box on the right highlights functionally anosmic individuals with thresholds above the range tested here, while the gray box on the left indicates highly sensitive normosmic individuals who responded YES to all odorant concentrations. The detection threshold (x-axis) is expressed in the base-10 logarithm of the nominal concentrations. For color figure refer to online version.