Subjective versus Psychophysical Measures of Chemosensory Alterations following COVID-19 Free

COVID-19 causes both short-term and long-term morbidity across various organ systems [1‒3], with chemosensory dysfunction remaining a hallmark feature in both acute and chronic COVID-19 [4, 5]. Early reports associating olfactory loss with acute COVID-19 [6, 7] have evolved into tracking the prevalence of persistent olfactory dysfunction (OD) [8]. Although OD has a diagnostic role for acute COVID-19, studies have reported varying estimates of sensitivity and specificity [9‒12]. Diagnosis of persistent OD is limited by variability in individual perception of olfactory function, but patient-reported, or subjective, olfactory and gustatory function remains the mainstay of chemosensory evaluation. Subjective function is easy to assess, time-efficient, and cost effective with options ranging from clinical interview to formal survey. However, psychophysical assessment of olfaction typically requires considerable resources: more time, equipment, and person power [13, 14].

Individual perception of chemosensory function is multifactorial: though many individuals correctly report the degree of olfactory and gustatory dysfunction they experience, certain factors make predicting the true degree of dysfunction difficult, including older age and cold symptoms [15]. Additionally, impairment in olfaction can easily be misattributed to gustatory dysfunction due to the perception of retronasal olfaction as gustation [16, 17]. Among patients with COVID-induced OD, the method of olfactory assessment accounted for notable differences in prevalence, where studies using subjective measures of olfactory function suggested a 44.4% prevalence of olfactory loss and those using psychophysical measures suggested a 76.7% prevalence [18]. These data propose that isolated subjective assessment may underestimate the true prevalence of OD in COVID-19 patients. Further complicating estimation of the prevalence of persistent OD following COVID-19 are the different types of OD: quantitative dysfunction or alteration in perceived strength of an odor, and qualitative dysfunction or alteration in perception of odor character [19]. Options for assessing qualitative olfactory function remain largely limited to subjective assessments, which have been shown to be particularly useful in assessing this type of OD [20].

Utility of psychophysical testing, like Sniffin’ Sticks, largely depends on its ability to provide more objective assessment of olfaction than widely used subjective measures [21]. The objectivity of various olfactory assessments has been previously examined in the literature. Mori et al. [22] examined the association between the open essence scent identification test card and other conventional olfactory assessments widely used in Japan, including visual analog scale (VAS), and found a statistically significant correlation among the varying tests.

Though numerous studies have attempted to evaluate olfactory loss as a diagnostic indicator of COVID-19 infection [9, 10], few have examined the association of subjective and psychophysical measures of olfactory function over time in this patient population. This study seeks to clarify the association between subjective and psychophysical measures of olfaction and gustation among individuals with self-reported persistent OD after COVID-19 infection.

Study participants were recruited by the CommonScents Research Group as part of an NIH-funded, prospective, longitudinal cohort study examining the relationship between chemosensation and neurocognition (NIH/NIDCD:K23DC019678-01). Participants entered the study following rhinology clinic referral on the basis of self-reported OD for more than 3 months, by volunteering through the online RecruitMe platform hosted by Columbia University Irving Medical Center (CUIMC), or by responding to posted flyers on the greater CUIMC campus. Informed consent was obtained from all participants. All recruitment and research practices were approved by the Institutional Review Board through the Human Research Protection Office at CUIMC (protocol AAAT6202).

This study included any adults with a personal history of COVID-19 confirmed by PCR testing or infection-specific serology. All participants had self-reported olfactory or gustatory dysfunction present for over 3 months. Participants with a history of preexisting OD, chronic rhinosinusitis, neurological disease, severe head trauma, a SNOT-22 rhinologic subdomain score of greater than 21, an answer of “severe” or worse for any SNOT-22 item, or an olfactory cleft endoscopy score (OCES) indicating probable rhinologic disease were excluded from this analysis [23].

Each participant completed a comprehensive survey, providing demographic information and sensory self-evaluation. Participants were asked to provide date of birth and sex assigned at birth. They were also asked to rate their senses of olfaction and gustation on a VAS from zero to 100 with zero as no sense of olfaction/gustation and 100 as an excellent sense of olfaction/gustation. Participants completed the survey portion of the study online utilizing the REDCap data management system prior to each in-person visit at zero and 1 year from initial intake, with select participants reporting for additional interval testing at 4 months (n = 36) and 8 months (n = 7) after baseline assessment. Sensory self-evaluation was completed prior to each round of in-person assessment.

After online intake, participants underwent in-person chemosensory psychophysical testing using Sniffin’ Sticks olfactory (Burghart Messtechnik GmbH, Holm, Germany) and Monell gustatory tests. Individuals attended initial baseline and follow-up visits at zero and 1 year from initial intake, with select participants reporting for additional interval testing at 4 months and 8 months after baseline assessment.

Participants underwent olfactory evaluation with Sniffin’ Sticks, a psychophysical clinical examination that assesses olfactory function via three distinct olfactory tests: threshold (T), discrimination (D), and identification (I) [13, 24]. Normosmia is defined as a total TDI score greater than 30.75 out of a total possible score of 48. Hyposmia is defined as less than or equal to 30.75 but greater than 16.25. Anosmia is any score less than or equal to 16 [25].

Gustatory evaluation was achieved with a brief gustation test developed at the Monell Chemical Sciences Center. Participants had to taste six different compounds twice for a total of 12 tastings and identify each via a forced-choice paradigm [14]. Performance was graded by the number of correct identifications out of a total possible score of 12.

Descriptive statistics were summarized as frequencies and proportions for categorical variables and means and standard deviations for continuous variables. Generalized estimating equation (GEE) models with exchangeable covariance matrices were utilized to analyze the association between predictors and outcomes in order to account for the correlation between subjects’ repeated measurements. For modeling, both predictors and outcomes were treated as continuous, using the normal distribution and the identity link function. The distribution of the residuals for each model was used to identify any potential outliers and assess model fit. If initial outliers were found, analyses were rerun after censoring outliers by setting them to missing. Models were adjusted for time (i.e., follow-up visit), age at baseline assessment, and sex at birth. The beta coefficient and 95% confidence interval of the predictor’s main effect were reported from both unadjusted and adjusted models, while the main effect of time was reported from the adjusted models. The predictor-by-time interaction terms were also evaluated but not reported unless found to be statistically significant. Partial R² values for the predictors were also evaluated for each model. Lastly, analyses of score components were also conducted if the overall score was found to be significant in the adjusted model. All analyses were carried out using SAS version 9.4. All statistical tests were two-sided and used a p < 0.05 to determine statistical significance.

Participant characteristics and demographics are shown in Table 1. The sample cohort (n = 65) is comprised predominantly of individuals experiencing persistent OD following COVID illness prior to vaccination availability and not requiring hospitalization during the initial phase of the pandemic in 2020. Timing of illness for this cohort closely follows emergence of the SARS-CoV-2 B-type virus and subsequent variants Alpha (B.1.1.7), Beta (B.1.351), and Delta (B.1.617.2) [26, 27]. Baseline chemosensory measures for the cohort are summarized in Table 2. Beta estimates from GEE analysis demonstrated a statistically significant association between subjective and psychophysical measures of olfaction. For each one-point increase in subjectively-reported olfactory ability, there is, on average, a 0.11 (95% CI: 0.06, 0.16; p < 0.001) point increase in TDI score while adjusting for age at baseline assessment, sex, and follow-up time. Additionally, subjectively-reported olfactory ability can explain 14% of the variance beyond that of time, age at baseline assessment, and sex assigned at birth for psychophysical olfactory score. Overall associations between subjective and psychophysical chemosensory measures are summarized in Table 3 and Figure 1. When looking at each individual component of the TDI score, there are statistically significant associations between both the discrimination and identification domains and subjectively-reported olfaction. For each one-point increase in subjectively-reported olfactory ability, there is, on average, a 0.04 (95% CI: 0.02, 0.06; p < 0.001) point and 0.05 (95% CI: 0.03, 0.07; p < 0.001) point increase in discrimination and identification scores, respectively, when adjusting for age at baseline assessment, sex, and follow-up time. Associations between subjective olfaction and each component of the TDI score are detailed in Table 4 and Figures 2-4.

Additionally, the main effect of time was found to be statistically significant when modeling the association between subjectively-reported olfactory ability and psychophysical olfactory score (AEst = 2.95, 95% CI: 1.00, 4.89; p = 0.003), and when modeling the association between subjectively-reported gustation and psychophysical olfactory score (AEst = 2.62, 95% CI: 0.66, 4.58; p = 0.009). Analysis of the interaction between the main effect of the predictor and the main effect of time was not statistically significant for any of the associations listed in Table 3.

Univariable analysis also showed a statistically significant association between subjectively-reported gustation and psychophysical olfactory score, but this association was no longer present after adjusting for age at baseline assessment, sex, and follow-up time. There was no statistically significant association between subjectively-reported gustation and psychophysical gustatory score or between psychophysical measures of olfaction and gustation, which remained unchanged when re-running analyses after censoring outliers.

The results of our study suggest an association between subjective report of olfactory function and the cumulative score from psychophysical testing. Additionally, the general subjective rating of olfactory function is associated with the discrimination and identification subdomains of the psychophysical testing, but not with the threshold subdomain. These findings demonstrate the utility of including patients’ subjective olfactory function within an overall evaluation; however, the inability to accurately self-report threshold may limit the utility of trending subjective olfactory function for recovery.

Similar to our results, a study conducted before the COVID-19 pandemic found that the rating of olfactory function on a VAS among patients with OD secondary to chronic rhinosinusitis, infection, or prior head trauma tracked with significant correlation to total Sniffin’ Sticks score (TDI) [28]. However, it was found that subjective VAS scores were significantly associated with all subdomains of the Sniffin’ Sticks assessment. In our study, subjective olfactory function was significantly associated with discrimination and identification subdomains, but not threshold. In a separate study looking specifically at patients with olfactory loss related to COVID-19, Bordin et al. [29] also found a significant correlation between VAS scores and Sniffin’ Sticks TDI scores at 6 months after their COVID diagnosis. These studies, along with our findings, suggest that subjective olfactory loss may be a reliable measure to provide a rough estimate of a person’s overall olfactory function. However, our study highlights a potential shortcoming of subjective olfactory report. It is also important to note that many psychophysical tests are predicated on odor recognition but lack the ability to classify olfactory threshold, highlighting a need for thoughtful selection of olfactory testing methods [30].

Prajapati et al. [31] found a correlation between subjective olfactory function and identification scores using the BSIT, and Jang et al. [15] found that the majority of people in their study were able to accurately assess their sense of olfaction with assessment of identification only. Our results similarly demonstrated a significant relationship between identification and subjective olfactory function; but, by including comprehensive psychophysical testing in our study design, we are able to highlight a specific caveat: that threshold is not associated with a person’s subjective reporting of their olfactory function in our post-COVID-19 patient population. There have been mixed results in the previous literature regarding the association between subjective olfactory scores and psychophysical threshold scores. Philpott et al. [32] found no correlation between threshold and subjective olfactory scores among patients with subjectively normal olfaction. The results of our study and from the prior literature demonstrate the importance of conducting full psychophysical olfactory testing in order to adequately assess threshold.

The neurocognitive underpinnings of olfaction remain relatively understudied [33, 34]. Prior research has shown significant cognitive influence on discrimination and identification domains of olfaction but not on threshold [35]. These findings may provide insight into why participants’ subjective assessments of olfaction are significantly correlated with only discrimination and identification. More research is needed to understand the relationship between olfaction and neurocognition and how that relationship may affect other cognitive domains.

The findings of this study showed no significant associations between subjective and psychophysical gustation, psychophysical olfaction and gustation, and subjective gustation and psychophysical olfaction. While patients often report new onset loss of olfaction and gustation with COVID-19 infection, the main chemosensory deficit may exist in the olfactory domain [36]. Olfaction and gustation are part of a collective chemosensory system that combines multimodal input into a unified perception of flavor, making it difficult for an individual to localize chemosensory loss to any one particular domain [37, 38]. Despite the lack of statistically significant findings in the current study while controlling for baseline age, sex, and follow-up time, there was a statistically significant association between subjective gustation and psychophysical olfaction when disregarding these covariates. It is possible that subjective gustatory loss can be explained by psychophysical OD; however, recent research suggests the presence of isolated gustatory deficits secondary to COVID-19 [39, 40]. Further work is needed to determine the relative contributions of the distinct chemosensory systems to the overall deficits in this population.

Studies examining the persistence of OD over time have shown mixed results: some studies suggest that over time individuals became less aware of their OD, while others suggest that subjective OD persists long beyond resolution of acute infection [29, 41‒43]. In this study, the main effect of time analysis suggests that individuals significantly improved in their psychophysical olfactory scores while adjusting for subjective olfactory scores, age at baseline, and sex. However, there were no statistically significant findings in analyzing the interaction between the predictor and the main effect of time as detailed in Table 3; thus, the results of this study cannot make any claims on whether the associations between subjective and psychophysical measures change over time. More research is needed to understand how insight into one’s psychophysical chemosensory scores may affect subsequent perception of olfactory loss, and what implications these factors have for future quality of life assessments. Although for recruitment purposes, inclusion criteria were set for individuals with at least 3 months of self-reported persistent OD, our population purposely skews toward a more chronic phenotype, as spontaneous recovery occurs at diminishing rates over the course of the initial year [8, 44].

The present study has limitations. Many of the participants had OD after infection with the B-type SARS-CoV-2 strain and early stage variants. Frequency and severity of chemosensory dysfunction secondary to later variants may differ from these initial strains [45‒47], but chemosensory changes remain a substantial issue across recent variants and subvariants as well as after vaccination [48, 49]. Additionally, unavailability of baseline chemosensory function prior to COVID-19 limits the true evaluation of chemosensory function over time; however, all participants met strict criteria for participation, including antecedent subjective changes in chemosensation. Finally, due to limited sample size, analysis may fail to show underlying associations.

In pursuit of future work, research assessing chemosensation is particularly needed in a racially and ethnically diverse population. This population is particularly important given the disproportionate effect that the COVID-19 pandemic has had on ethnic and racial minorities [50]. Also, in order to inform clinical practice and future psychophysical testing design, further research is needed to elucidate specific factors that contribute to differences between subjective and psychophysical measures of chemosensation. Finally, commitment of chemosensation researchers to a standardization of terminology used to describe OD would be beneficial in streamlining the literature and providing a framework for future scoping and systematic reviews [51].

This study supports the overall association between subjective perception of chemosensory functioning and more comprehensive psychophysical testing results. Utilizing psychophysical chemosensory testing inclusive of threshold assessment is necessary to facilitate recovery trending and for the assessment of true chemosensory function in patients with COVID-19-associated chemosensory dysfunction.

We would like to thank the NIDCD for supporting this research. Study data were collected and managed using REDCap electronic data capture tools hosted at Columbia University [52, 53]. REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing (1) an intuitive interface for validated data capture; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for data integration and interoperability with external sources.

This study protocol was reviewed and approved by the Institutional Review Board through the Human Research Protection Office at CUIMC, protocol number AAAT6202. Written informed consent was obtained from participants through REDCap [54].

The authors have no conflicts of interest to declare.

This project was funded by the NIDCD of the National Institutes of Health under grant K23DC019678-01. This publication was supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant No. UL1TR001873. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

B.J.V. and P.T.J. coordinated the study, performed data collection, and wrote the manuscript. C.J.S. and T.-H.C. performed statistical analyses and wrote the statistical methods. L.W.G., F.F.C., J.P.T., and J.B.G. conceptualized the project, led initial data collection, and reviewed the manuscript. T.M.S. continued data collection and reviewed the manuscript. D.A.G. and P.V.J. reviewed and revised the manuscript. T.E.G. and D.P.D. developed methodology, oversaw the project, and reviewed the manuscript. J.B.O. led, managed, and coordinated the project and wrote the manuscript.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.