Recovery of Parosmia Preferentially Influences Longitudinal Improvement of Quality of Life in C19OD

ABSTRACT

Background

Persistent COVID‐associated olfactory dysfunction (C19OD) negatively impacts quality of life (QoL). This prospective longitudinal cohort study sought to determine which aspects of chemosensory recovery may preferentially influence QoL improvement in this population.

Methods

Individuals with C19OD (N = 100) completed chemosensory and QoL assessment with Sniffin’ Sticks, Taste Assessment, Questionnaire of Olfactory Disorders‐Negative Statements (QOD‐NS), and QOD‐PAR at baseline and 1 year. Multivariable analyses assessed baseline and longitudinal associations between chemosensory dysfunction and QOD‐NS.

Results

QoL associated with TDI (Coef.; [CI]; p‐value: −0.32; [−0.62, −0.032]; 0.030), threshold (−0.82; [−1.5, −0.14]; 0.019), QOD‐Par (−1.0; [−0.20, −1.9]; 0.016), and self‐reported gustatory dysfunction (GD) (−7.4; [−14, −0.46]; 0.037) at baseline assessment, though quantitative GD did not associate with QoL. Longitudinally, improvements in discrimination (−1.2; [−2.3, −0.18]; 0.023) and QOD‐Par (−2.5; [−0.99, −4.1]; 0.002) scores were associated with improved QoL. Individuals with improved parosmia symptoms experienced a 12.86 (12.99, 0.040) QOD‐NS point improvement from baseline compared to 5.91 (7.98, < 0.001) for those with persistent parosmia.

Conclusions

Independent of mental health, chemosensory function independently drives QoL in C19OD. Parosmia, low odor threshold, and patient‐reported GD are all associated with poorer QoL at baseline. Longitudinally, the ability to differentiate among odors is an important correlate of improvement of smell‐related QoL in C19OD patients, while parosmia is the strongest driver of longitudinal recovery of smell‐related QoL in C19OD. Trending domain‐specific olfactory performance in C19OD may allow for improved patient counseling and QoL prognosis.

1. Background

Olfactory dysfunction (OD) has long been associated with a diminished quality of life (QoL) [1, 2], negative health outcomes such as cognitive impairment and neurodegeneration [3, 4], reduced environmental safety awareness [5], and development of mental health comorbidities [6, 7], all of which have become more pronounced in the setting of persistent COVID‐associated olfactory dysfunction (C19OD) [8, 9, 10, 11]. C19OD is a key feature of “long COVID,” a condition impacting 6.9% of adults and defined by persistent COVID symptoms lasting more than 3 months following diagnosis [12, 13, 14]. Nearly a third of all patients who recover from COVID‐19 experience persistent qualitative chemosensory dysfunction (phantosmia, parosmia, phantogeusia, parogeusia) and a quarter of patients present with quantitative chemosensory dysfunction (hyposmia, anosmia, hypogeusia, ageusia) months after initial infection [13]. Among those with persistent C19OD, perceived OD and presence of parosmia have emerged as more notable indicators of poor QoL outcomes compared to quantitative psychophysical OD [8, 15, 16, 17]. Studies examining how deficits in specific chemosensory domains affect longitudinal QoL outcomes are needed to improve clinical interventions.

Quantitative OD is often captured using core psychophysical domains of odor threshold (sensitivity to detect odor concentrations), odor discrimination (ability to differentiate among odors), and odor identification (correctly corresponding an odor with its name), while qualitative OD, particularly parosmia, is conventionally measured with validated questionnaires [18] despite reports of standardized measures of valence perception [19]. Psychophysical gustatory function is typically assessed via the ability to correctly perceive and name five taste stimuli: salty, sweet, bitter, sour, and umami, while qualitative gustatory dysfunction (GD) often relies on self‐reported altered sense of taste. Notably, many patients self‐report GD in C19OD [20], where much of this perceived taste loss can be attributed to disruption of retronasal olfaction and flavor recognition [21], as true GD is rare in this population [22, 23, 24]. This semantic discrepancy between the academic definition and the general public's interpretation of taste and flavor prompts further characterization of how perceived GD may also impact QoL among those with concomitant C19OD.

Although persistent C19OD and long COVID similarly result in worse QoL outcomes [25, 26], there are no studies that investigate which domains of psychophysical chemosensation are most critical for longitudinal improvement of QoL in C19OD. Thus, in this prospective longitudinal cohort study, we sought to (1) build upon previous studies examining which aspects of chemosensation are most strongly associated with smell‐related QoL outcomes at baseline assessment and (2) investigate how the longitudinal evolution of specific domains of chemosensation in C19OD may impact QoL improvement over time.

2. Methods

2.1. Recruitment and Inclusion Criteria

A total of 100 individuals were enrolled in the study for baseline questionnaires and psychophysical olfactory assessment, with 46 participants returning for longitudinal follow‐up. Individuals were recruited and evaluated in accordance with proposed protocols approved by the Columbia University Irving Medical Center Institutional Review Board, via referral from clinicians for COVID‐related smell loss. Participants were additionally recruited from an online research recruitment platform affiliated with the Columbia University Irving Medical Center. All interested individuals provided informed consent prior to participation in the study.

Participants were eligible based on the following inclusion criteria: age ≥ 18 years, confirmed SARS‐CoV‐2 positivity via PCR or SARS‐CoV‐2 nucleocapsid antibody serology (or clinical diagnosis of COVID‐19 if diagnosed prior to widespread testing availability), and self‐reported persistent OD (> 3 months). Exclusion criteria included any diagnosed pre‐existing OD, SNOT‐22 rhinologic subdomain score ≥ 21 as assessed on pre‐evaluation surveys, any pre‐existing neurologic or other health problems that can independently cause OD, and any self‐reported pre‐existing functional cognitive deficit.

2.2. Validated Questionnaires and Chemosensory Assessment

Prior to in‐person chemosensory evaluation, participants completed surveys and validated questionnaires assessing patient demographics, COVID‐19 history, perception of COVID‐19 illness severity, perceived smell/taste dysfunction, parosmia status, and QoL. The Questionnaire of Olfactory Disorders‐Negative Statements (QOD‐NS) [27, 28] was utilized to assess olfactory‐related QoL. Higher scores indicate a poorer smell‐related QoL, with 12.5 representing a cutoff for normal versus abnormal scores [29]. The QOD‐Par, reflective of the Landis questionnaire [18], was used to quantify parosmia numerically, where lower scores correspond to a more severe level of parosmia (Table S1), the inverse of QOD‐NS. Further, participants were stratified by categorical parosmia status at each time point using survey questions as outlined in Table S1. In addition, participants completed the Self‐Administered Comorbidity Questionnaire (SCQ) [30], Beck Anxiety Index (BAI) [31, 32], and Patient Health Questionnaire‐9 (PHQ‐9) [33] to provide physical and mental health variables for use in statistical analysis.

A trained research assistant administered the validated Odofin Sniffin’ Sticks test (Burghart Messtechnik GmBH, Holm, Germany), including odor threshold (T), odor discrimination (D), and odor identification (I), to all participants for olfactory assessment at baseline and follow‐up. Measurements of T, D, and I, and the combined TDI outcome, were completed following the protocol popularized by Hummel et al. (1997) [34] in the same manner described in other studies [9]. Individual scores for T, D, and I range from 1 to 16, with the sum of the individual scores representing the overall TDI score. TDI score allows for stratification of individuals into olfactory function categories of normosmic (30.75 or higher), hyposmic (scores between 16.25 and 30.5 points), and anosmic (score of 16.5 or less) [35].

Quantitative taste assessment was completed using a protocol outlined by Douglas et al. (2018) [36]. Six taste solutions were prepared, including denatonium benzoate (bitter), PTC (bitter), quinine (bitter), sodium chloride (salty), sucrose (sweet), and distilled water (no flavor). Participants were presented with each tastant twice and instructed to deem each stimulus as “salty,” “sweet,” “bitter,” “sour,” or “no flavor.” Scoring on the quantitative taste assessment ranges from 0 to 12, with each correct response corresponding to one point, such that a higher score indicates intact gustatory function. Perceived GD was evaluated by the presence of self‐reported taste dysfunction using the question provided in Table S2.

2.3. Statistical Analysis

While the specific line of questioning evaluated in this investigation precluded an anchored power analysis, to guide study composition, we planned for an analysis including seven predictor variables. Using a G*Power estimate, we arrived at a target sample size of approximately 100 participants required to detect a moderate effect size. Given the nature of C19OD, we sought to include participants beyond this number, depending on participant availability, with eligibility and loss to follow‐up important considerations in the target sample size.

IBM SPSS Statistics for macOS, version 29.0 (IBM Corp., Armonk, NY, USA) was used for all statistical analyses. To account for any potential bias in longitudinal analysis due to loss to follow‐up, participants were stratified by follow‐up status, and a comparison of baseline characteristics, including age, sex, education status, comorbidity index, PHQ‐9, BAI, and all primary outcome variables, including chemosensory measures and QOD‐NS, was completed using independent samples t‐tests.

Chemosensory measures, including overall TDI, odor threshold, odor discrimination, odor identification, QOD‐Par, self‐reported gustation status, and psychophysical gustation assessment, were compared to QOD‐NS score at baseline using multivariable linear regression inclusive of age, sex, comorbidity index, BAI, and PHQ‐9 scores as covariates. Among those who completed follow‐up assessment, longitudinal change scores were developed for each chemosensory measure and QOD‐NS score. Change in each chemosensory measure was compared to change in QOD‐NS score over time using multivariable linear regression with the same covariates. To assess for potential shared variance among chemosensory predictors at baseline and longitudinally, Pearson's correlations were computed across all baseline measures and across all longitudinal change variables. Further, participants completing assessments at both time points were stratified into groups (1) improved parosmia (denoting a recovery of parosmia status at follow‐up as compared to baseline) and (2) consistent parosmia (meeting the above parosmia criteria at both time points). Paired samples t‐test was used to evaluate change in QOD‐NS score over time within each group, and independent samples t‐test was utilized to compare QOD‐NS change scores between groups. For all chemosensory measures, any inverse relationship with QOD‐NS in the regression models would suggest that better performance on the chemosensory measure at baseline relates to higher QoL. This pattern also holds true for the longitudinal analysis, where any inverse relationship with the QOD‐NS change score would indicate that improvement on the chemosensory measure corresponds to an improvement in QoL over time.

3. Results

3.1. Cohort Demographics

Baseline surveys, QOD‐NS, QOD‐Par, and psychophysical olfactory assessments were performed for 100 participants, of which 78 (78%) were female and 22 (22%) were male (Table 1). Of these, 46 participants returned for longitudinal follow‐up, including 32 (70%) females and 14 (30%) males. Baseline testing was performed at a mean (SD) 474 (312) days from COVID‐19 diagnosis, while longitudinal follow‐up testing was performed at a mean (SD) 887 (255) days from COVID‐19 diagnosis. Among the study population, 16% reported having ever smoked. Most participants reported perceived severity of their preceding COVID‐19 illness as substantially less (28%) or slightly less (18%–26%) severe than others. The mean (SD) age of participants was 42.3 (14.7) years old at baseline testing and 40.4 (13.3) years old at follow‐up.

TABLE 1.

Demographics and descriptive statistics.

	Baseline, N = 100 a	Follow‐up, N = 46 a
Age	42.3 (14.7)	40.4 (13.3)
Sex
Female	78 (78%)	32 (70%)
Days from COVID‐19 diagnosis	474 (312)	887 (255)
Comorbidity index	2.9 (3.3)	2.4 (2.8)
PHQ‐9 score	5.7 (6.5)	5.5 (6.4)
BAI score	8.0 (9.8)	6.5 (8.2)
Ever smoked
Yes	16 (16%)	7 (15%)
Perceived severity of COVID illness
Substantially less severe	28 (28%)	13 (28%)
Slightly less severe	18 (18%)	12 (26%)
About the same as others	17 (17%)	9 (20%)
Slightly more severe	17 (17%)	6 (13%)
Substantially more severe	6 (6%)	1 (2%)
OD status
Normosmia	26 (26%)	19 (41%)
Hyposmia	60 (60%)	25 (54%)
Anosmia	14 (14%)	2 (4%)
TDI score	25.4 (7.8)	28.7 (6.8)
QOD‐NS score	25.8 (12.3)	19.5 (12.4)
QOD‐Par score	9.0 (2.7)	8.0 (2.5)

^a Mean (SD); n (%).

While not all baseline participants returned for follow‐up, comparative analysis revealed few significant differences between the retained longitudinal cohort and those lost to follow‐up based on age (p = 0.3), sex (p = 0.060), education status (p = 0.7), QOD‐NS score (p = 0.5), TDI score (p = 0.5), odor threshold (p = 0.7), odor discrimination (p = 0.3), odor identification (p = 0.8), QOD‐Par (p = 0.7), quantitative taste assessment (p = 0.9), self‐reported GD status (p = 0.4), comorbidity index score (p = 0.4), PHQ‐9 (p = 0.8), or BAI (p = 0.7) upon performing Pearson's chi‐squared (education status and self‐reported GD) and independent samples t‐tests (all other variables) (Table S3).

3.2. Chemosensory Function and Overall QoL

The majority of the study population was found to have psychophysical OD (anosmia or hyposmia) at both time points, with 74 (74%) of participants presenting with OD at baseline and 27 (58%) at follow‐up (Table 1). The mean (SD) corresponding TDI scores for these time points were 25.4 (7.8) and 28.7 (6.8), respectively, both within the range for hyposmia. At both time points, mean QOD‐NS scores were found to be in the abnormal range with mean (SD) scores of 25.8 (12.3) at baseline and 19.5 (12.4) at follow‐up. QOD‐Par was 9.0 (2.7) at baseline and 8.0 (2.5) at follow‐up.

3.3. Association of Chemosensory Measures and QoL at Baseline Assessment

At baseline assessment, when controlling for covariates of age, sex, comorbidity index, BAI, and PHQ‐9 score, we noted that overall TDI score (Coef.; [CI]; p‐value: −0.31; [−0.62, −0.032]; 0.030), odor threshold score (−0.82; [−1.5, −0.14]; 0.019), QOD‐Par score (−1.0; [−0.2, −1.9]; 0.016), and self‐reported GD score (−7.4; [−14, −0.46]; 0.037) were all associated with poorer QoL as evidenced by higher QOD‐NS score (Figure 1a and Table S4). Odor discrimination (−0.47; [−1.2, 0.28]; 0.2), odor identification (−0.61; [−1.3, 0.12]; 0.10), and taste assessment (−0.32; [−2.2, 1.6]; 0.7) did not associate with QOD‐NS score variations at baseline assessment.

FIGURE 1.

(a) Baseline and (b) longitudinal predictors of QoL status and improvement.

3.4. Association of Changing Chemosensory Measures on Longitudinal QoL Evolution

When controlling for covariates of age, sex, comorbidity index, BAI, and PHQ‐9 score, we observed that a change in odor discrimination (−1.2; [−2.3, −0.18]; 0.023) and QOD‐Par score (−2.5; [−0.99, −4.1]; 0.002) were associated with change in QOD‐NS score, such that improvement in any of these chemosensory modalities corresponded to an improvement in QoL over time (Figure 1b and Table S5). Change in TDI score (−0.43; [−0.98, 0.13]; 0.13), odor threshold (0.32; [−0.69, 1.3]; 0.5), odor identification (−0.94; [−2.1, 0.19]; 0.10), self‐reported GD (5.1; [−3.1, 13]; 0.2), or quantitative taste assessment (0.18; [−2.7, 3.0]; 0.9) did not correspond to a significant change in QOD‐NS score.

When stratifying all participants presenting for follow‐up parosmia change status, both participants in the improved parosmia group (baseline mean [SD], follow‐up mean [SD], p‐value: 23.0 [10.3], 10.1 [11.9], 0.040) and the consistent parosmia group (27.4 [11.1], 21.3 [11.6], < 0.001) demonstrated improvement in QOD‐NS score (Figure 2 and Table S6) at follow‐up. However, the QoL improvement effect size was significantly larger for the group that experienced resolution of parosmia at follow‐up (mean (SD): 12.9 [13]) versus the group experienced consistent parosmia at both time points (6.2 [7.8]) (p = 0.039), resulting in a clinically significant difference in mean values between baseline and follow‐up in the improved parosmia group.

FIGURE 2.

Change in QOD‐NS score over time by parosmia status change over time.

3.5. Analysis for Potential Shared Variance Among Chemosensory Predictors

The three quantitative olfactory subtests, odor threshold, odor discrimination, and identification, were all moderately intercorrelated (r = 0.46–0.55, p < 0.001) and each was highly correlated with the composite TDI score (r = 0.81–0.82, p < 0.001) at baseline, reflecting that TDI is the sum of these components (Table S6). QOD‐Par was also moderately correlated with self‐reported GD status (r = 0.35, p < 0.001). Correlations among the other chemosensory baseline variables were generally weak (r < 0.30), indicating minimal overlap and allowing for interpretation with minimal collinearity. For the longitudinal analysis, the T, D, and I subdomain change scores were all moderately correlated with the composite TDI change score (r = 0.59–0.70, p < 0.001) (Table S7). Correlations among all other chemosensory variable change scores, including among T, D, and I subdomains, were weak (r < 0.30), allowing for interpretation of the longitudinal predictors on QOD‐NS change with minimal collinearity.

4. Discussion

In this prospective longitudinal cohort study, we (1) evaluated features of chemosensation associated with poorer smell‐related QoL in C19OD and (2) investigated how recovery of specific domains of chemosensation may impact QoL improvement over time. Our results suggest that the main factors negatively impacting QoL at baseline include deficits in overall TDI and threshold, as well as severity of parosmia and perceived taste dysfunction. Resolution of parosmia and, to a lesser degree, recovery of odor discrimination are the factors that have a significant impact on QoL improvement in C19OD. Notably, these associations exist independently of mental or physical health status, suggesting that chemosensory function is a critical aspect of long COVID to target for QoL improvement, and domain‐specific olfactory evaluation may inform clinical QoL prognosis.

In general, odor threshold is the domain of psychophysical olfactory assessment initially most strongly impacted by SARS‐CoV‐2 infection [37, 38]; however, there is no consensus in the literature on which domains regain function first in the recovery period. Some studies suggest that suprathreshold, centrally mediated domains (discrimination and identification) improve faster [37, 39], while others demonstrate that odor threshold may recover faster in line with its principal measurement of peripheral function [38, 40]. Our findings show a baseline association of overall TDI and odor threshold with diminished QoL, which is reasonable given the stronger initial insult to odor threshold following infection. Interestingly, the present study demonstrates that improvement in odor discrimination—not threshold—is a marker of QoL recovery in this population. It is not readily apparent why odor discrimination uniquely correlates with QoL improvement among quantitative olfactory domains, but it may be related to experiences of patients with parosmia who regain the ability to properly distinguish among different odors. It is possible that improvements in odor discrimination represent a point in recovery where participants have regained enough threshold ability to be able to detect suprathreshold scents and may be able to differentiate among different olfactory stimuli, even if the odors are not yet “smelling as they should” or odor identification is still difficult.

The practice of olfactory training is commonly recommended as a treatment option for patients with C19OD, and it inherently relies on odor discrimination function [41]. Further, the positive impacts offered by smell training are much more pronounced in suprathreshold domains of olfaction, including odor discrimination [42]. Given the QoL benefit that may be offered by improvement of odor discrimination, results from this study suggest that C19OD patients should continue to engage with smell training, particularly as it is a low‐burden, noninvasive treatment option. With the understanding that improvements in odor discrimination are a potential marker of clinical QoL improvement in this population, further studies tailoring smell training and other novel therapeutic modalities to target and enhance odor discrimination abilities are warranted.

Importantly, the minimal clinically important difference (MCID) of TDI based on prior literature is 5.5, odor threshold is 2.5, and discrimination and identification both have an MCID of 3 [43]. The MCID of QOD‐NS has been reported as 5.2 [29]. Thus, applying these pre‐determined values to the regression models reported in the results, MCIDs in TDI, odor threshold, odor discrimination, and odor identification do not correspond to clinically important differences in QOD‐NS scores at baseline. For the longitudinal regression model, a 3‐point improvement in discrimination longitudinally corresponds to a 3.6‐point improvement in QOD‐NS score over time, which is also not clinically significant. Thus, though low TDI and low odor threshold correspond to poorer QoL at baseline and improvements in odor discrimination longitudinally correlate with improvements in QOD‐NS over time, these changes may not be clinically appreciated by the patient or clinician.

Parosmia, however, clearly plays a large role in mediating QoL outcomes in C19OD, both at baseline and in predicting recovery. Other studies have demonstrated the important role of qualitative OD in C19OD, with the presence of parosmia [17] and perceived OD [16] seemingly impacting QoL more strongly than quantitative OD. These associations support our finding that improvement in parosmia may direct longitudinal QoL recovery. Importantly, statistically and clinically significant improvements in QoL were observed across the study cohort, regardless of parosmia improvement status. Compared to those with persistent parosmia, those who experienced parosmia resolution demonstrated a clinically significant change in QOD‐NS score, with the group mean below the 12.5 score cutoff considered to be within the “normal range,” indicating parosmia resolution. Further, the improved parosmia status group experienced a statistically larger effect size in QoL improvement than the persistent parosmia group.

When applying the longitudinal regression model, we can see that just a 3‐point improvement on QOD‐Par corresponds to an improvement in QOD‐NS score of 7.5, well above the MCID for QOD‐NS. A 3‐point improvement on QOD‐Par would, for example, correspond to a participant experiencing a 1‐point improvement on three of the four measures, such as going from “always” to “often,” “often” to “rarely,” or “rarely” to “never” regarding various questions aiming to quantify the pervasiveness of various domains of parosmia in the participant's daily life (Table S1). Taken together, we assert that improvements in parosmia both statistically and clinically correspond to improvements in smell‐related QoL in C19OD.

It is not surprising that parosmia relates so closely to QoL at baseline and in driving longitudinal QoL recovery, given the importance of retronasal olfaction in the experience of food [44]. While perceived GD was associated with poor QoL at baseline, psychophysical taste function performance was not. The colloquial experience of taste often includes retronasal olfaction [45, 46], which is crucial for the perception of flavor. Dysfunctional retronasal olfaction can contribute to parosmia and flavor disruption [47], while true GD is quite rare in this population [22, 23, 24, 48]. The lack of GD was also found in the present population, where we observed some overlap between perceived GD and QOD‐Par, as evidenced by a significant, moderate correlation between the variables at baseline (Table S6). Likely due in part to retronasal OD, C19OD is associated with shifting food consumption habits, often toward increased intake of highly processed foods [49], and has also been associated with increases in BMI [50]. The proper experience of food, beyond its importance for nutrition and physical health, is a key aspect of social functioning, mental health, and QoL [51]. Thus, it is reasonable that recovery in parosmia, particularly parosmia impacted by diminished retronasal olfaction, leads to large improvements in QoL.

A discussion of potential shared variance and collinearity among the various chemosensory predictors utilized across the regression models for baseline and longitudinal associations is warranted, given that all predictors are measuring some aspect of chemosensation. As TDI is a composite score of odor threshold, discrimination, and identification, baseline TDI scores are strongly correlated with its subdomain scores. Thus, the effects of TDI on olfactory‐related QoL at baseline likely represent that combined influence of these subdomains. Further, the moderate associations among the olfactory subdomains at baseline indicated that these components are related yet distinct dimensions of chemosensory performance. Thus, the associations between the individual olfactory measures and QoL outcomes at baseline likely reflect overlapping but non‐identical aspects of olfactory function. As previously addressed, there is overlap between perceived GD and QOD‐Par at baseline; thus, the significant relationship between perceived GD and QOD‐NS at baseline may be due in part to the colloquial experience of parosmia, particularly in the absence of psychophysical taste assessment correlation with QOD‐NS. Importantly, the baseline association between QOD‐Par and QOD‐NS is likely not strongly influenced by the other measures, given weak correlations between QOD‐Par and other chemosensory variables. The longitudinal change variables are less susceptible to possible shared variance and collinearity concerns, where there are no strong correlations noted among any of the chemosensory variable change scores. Moderate correlations exist between the TDI change score and the individual subdomain score as expected, though all other associations, including among the quantitative olfactory subdomain change scores, are weak. While there are overlapping features of many of these change variables, given that they are all chemosensory measures, we can conclude that the observed QOD‐NS improvements due to odor discrimination and parosmia improvements are largely independent of any changes occurring in the other chemosensory variables.

This study benefited from the use of validated measures for parosmia and QoL assessment in addition to validated psychophysical olfactory assessment using the extended Sniffin’ Sticks battery to encompass several smell modalities. Further, our use of psychophysical olfactory assessment to determine OD rather than solely relying on self‐reported smell loss was a strength of the study. In addition, we included only individuals who were self‐reporting chemosensory dysfunction, further strengthening observed relationships between psychophysical olfactory performance and QOD‐NS. In addition, longitudinal assessment of outcome measures allowed for an analysis investigating chemosensory predictors of QoL improvement and recovery, a novel aspect of this study. Lastly, the use of several covariates known to impact QoL outcomes, such as mental and physical health, strengthened the observed associations between chemosensation and QoL.

Our study also has limitations. While our overall study group was comprised of 100 individuals, we were significantly limited by loss to follow‐up due to challenges of retention in the post‐COVID era with a prospective cohort study of this design. To mitigate any potential biases introduced by this, we conducted an analysis comparing baseline results for various demographic factors and outcomes relevant to our study, stratified by follow‐up status (Table S3). None of the demographic, chemosensory, or outcome variables were found to be significantly different by follow‐up status. Since those who were lost to follow‐up were similar across most measures to those who returned for both assessments, we concluded that the results obtained in longitudinal assessment are likely representative of our full study population. Conclusions from this investigation should be interpreted as offering preliminary insights, where the data collected in this study can be used to inform formal power analysis for larger studies in the future. Demographically, our study population had an underrepresentation of males as compared to females, consistent with recruitment patterns in chemosensory research reported in the literature [52]. While this study was a prospective longitudinal study primarily aimed at understanding chemosensory predictors for QoL improvement in C19OD, it would have been ideal to have a normative non‐COVID control group; however, this was not possible given the study design, available resources, and difficulty in confirming lack of SARS‐CoV‐2 exposure.

In conclusion, our current study demonstrates that while deficits in overall TDI, threshold, presence of parosmia, and self‐reported GD are associated with poorer QoL in C19OD at baseline, improvements in odor discrimination and resolution of parosmia are the key aspects of chemosensation in driving QoL recovery in this population. Resolution of parosmia is likely the most important chemosensory factor driving clinically significant QoL recovery over time, while improvements in odor discrimination likely contribute to a lesser degree and may not be clinically appreciated by the patient. Importantly, these relationships persist independent of mental or physical health, further solidifying the critical role of chemosensory function for QoL in C19OD. Trending domain‐specific quantitative and qualitative olfactory performance in C19OD may allow for improved patient counseling and QoL prognosis in long COVID. Since domain‐specific olfactory improvement offers clinically significant QoL outcomes, development of novel treatments for olfactory recovery should be targeted at improving odor discrimination and resolving parosmia. Clinicians can deliver a message of optimism to C19OD patients regarding QoL improvement over time, regardless of parosmia status, particularly if there is recovery of odor discrimination function and parosmia.

Author Contributions

Conceptualization, methodology, and writing – review and editing: Tiana M. Saak and Jonathan B. Overdevest. Formal analysis, methodology, and writing – review and editing: Tiana M. Saak. Investigation, methodology, and writing – review and editing: Tiana M. Saak, Jeremy P. Tervo, Brandon J. Vilarello, Patricia T. Jacobson, Francesco F. Caruana, and Liam W. Gallagher. Validation, methodology, and writing – review and editing: Tiana M. Saak, Jeremy P. Tervo, Brandon J. Vilarello, Patricia T. Jacobson, Francesco F. Caruana, and Liam W. Gallagher. Data curation, methodology, and writing – review and editing: Tiana M. Saak, Jeremy P. Tervo, Brandon J. Vilarello, Patricia T. Jacobson, Francesco F. Caruana, and Liam W. Gallagher. Writing – original draft preparation, methodology, and writing – review and editing: Tiana M. Saak. Visualization, methodology, and writing – review and editing: Tiana M. Saak and Jonathan B. Overdevest. Supervision, methodology, and writing – review and editing: David A. Gudis, Davangere P. Devanand, and Jonathan B. Overdevest. Project administration, methodology, and writing – review and editing: Jonathan B. Overdevest. Funding acquisition, methodology, and writing – review and editing: Jonathan B. Overdevest.

Funding

This research was funded by the National Institute on Deafness and Other Communication Disorders, part of the National Institutes of Health, under grant number K23DC019678.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Survey questions utilized in QOD‐Par outcome at baseline and follow‐up from the Landis questionnaire.

Table S2: Question used to determine presence of perceived gustatory dysfunction at baseline and follow‐up.

Table S3: Characteristics and baseline scores of those who completed assessments at both time points (longitudinal cohort) versus those who were lost to follow up.

Table S4: Multivariate linear regression of chemosensory measures and QOD‐NS score at baseline assessment.

Table S5: Multivariate linear regression of chemosensory measure change and QOD‐NS score change over time.

Table S6: Paired analysis of QOD‐NS score over time with groups stratified by parosmia status change over time (including only participants with parosmia at baseline).

Table S6: Pearson's correlations among baseline chemosensory measures to assess for potential collinearity

Table S7: Pearson's correlations among longitudinal chemosensory measure change scores to assess for potential collinearity