COVID-19 Related Chemosensory Changes in Individuals with Self-Reported Obesity

Background/objectives: Individuals with obesity show alterations in smell and taste abilities. Smell and taste loss are also the most prominent neurological symptoms of COVID-19, yet how chemosensory ability present in individuals with obesity with a positive COVID-19 diagnosis is unknown. Subjects/Methods: In this secondary analysis of a cross-sectional global dataset, we compared self-reported chemosensory ability in participants with a respiratory illness reporting a positive (C19+; n = 5156) or a negative (C19−; n = 659) COVID-19 laboratory test outcome, who also self-reported to be obese (C19+; n = 433, C19−; n = 86) or non-obese. Results: Compared to the C19− group, C19+ exhibited a greater decline in smell, taste, and chemesthesis during illness, though these symptoms did not differ between participants with obesity and without obesity. In 68% of participants who reported recovery from respiratory illness symptoms (n=3431 C19+ and n= 539 C19−), post-recovery chemosensory perception did not differ in C19+ and C19− diagnosis, and by self-reported obesity. Finally, we found that all chemosensory and other symptoms combined predicted the C19+ diagnosis in participants with obesity with a moderately good estimate (63% accuracy). However, in C19+ participants with obesity, we observed a greater relative prevalence of non-chemosensory symptoms, including respiratory as respiratory and GI symptoms. Conclusions: We conclude that despite a presumed lower sensitivity to chemosensory stimuli, COVID-19 respondents with obesity experience a similar self-reported chemosensory loss as those without obesity, and in both groups self-reported chemosensory symptoms are similarly predictive of COVID-19.


INTRODUCTION 70
According to the World Health Organization, globally 13% of adults aged 18 years and 71 over reported to have obesity in 2016 (1). Within the context of the ongoing COVID-19 72 pandemic, intriguingly, countries with the highest prevalence of obesity also recorded a high 73 death rate from COVID-19 infection (2). Although, an increased in susceptibility to viral 74 infection with obesity is unknown, a recent review by Stefan et al. concluded that obesity is a 75 strong and independent determinant of morbidity and mortality in patients infected with  CoV-2, the virus responsible for . A recent analysis also indicated that 77 COVID-19 mortality in patients with obesity is higher than that of other comorbidities, including 78 diabetes, hypertension, asthma, and cancer (4). In addition to greater risk for COVID-19 related 79 poor health outcomes (3,5,6), patients with obesity are more likely to require hospitalization, 80 especially in young adults with a Body Mass Index (BMI) >30 kg/m (7). Overall, current 81 evidence suggests that obesity significantly interacts with the pathogenesis of  Despite this risk, COVID-19 chemosensory symptoms have not yet been systematically assessed 83 in this population group. 84 Disturbances in smell and taste emerged as a predominant neurological symptom of 85 COVID-19 infection, with 77 percent of COVID-19 patients reporting sudden olfactory and 86 gustatory dysfunctions in a recent meta-analysis (8). In a recent analysis, we also reported 87 quantified smell loss as the best predictor of COVID-19, compared to other common non-88 chemosensory symptoms (9). However, these studies did not delineate COVID-19-related 89 chemosensory impact in patients with obesity. This is especially important because individuals 90 with obesity typically have existing lower taste sensitivity, and lower capacity to detect and 91 identify odors than individuals without obesity (10). Particularly, excessive body weight has 92 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 been shown to be associated with impaired taste for sweet and salty foods, alteration in fat/fatty 93 acid-sensing, reduced ability to identify correct taste (11)(12)(13) in taste detection thresholds (11-94 15). These obesity-related chemosensory dysfunctions are driven by production of pro-95 inflammatory factors from adipose tissues, leading to impairment in olfactory receptors (16) and 96 a decline in taste bud and taste progenitor cells (17,18), respectively. Considering that marked 97 inflammation with obesity also seems to favor viral infections (19,20), the interaction between 98 existing chemosensory deficiency in adults with obesity and COVID-19 related chemosensory 99 impairments are unknown. 100 Existing gustatory and olfactory sensory deficiency due to obesity may mask the viral-101 induced diminished taste and smell self-reported experiences, leading to a higher portion of 102 undetected cases in this population (21), we need a better understanding of how chemosensory 103 profile changes in patients with obesity. Furthermore, in light of the potential for using oro-naso 104 sensory perception as an early marker of SARS-CoV-2 infection (22)(23)(24), it needs to be assessed 105 whether the predictive relation between the chemosensory loss and COVID-19 illness 106 generalizes to participants with obesity. With continually increasing death rates projected well 107 into 2021 using a second statistical model (25), and a wave of infections sweeping through 108 countries worldwide, it is imperative to understand the impact of SARS-COV-2 virus on 109 chemosensory dysfunction in COVID-19 patients in the high-risk category such as populations 110 with excessive body weight. Here, we systematically describe and compare the chemosensory 111 perception (smell, taste, and chemesthesis) and related symptomatology in COVID-19 in non-112 hospitalized adults with or without self-reported obesity. Following our pre-registered analysis 113 plan (26), we hypothesized that the participants with obesity will report less smell loss during 114 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Study Design 119
To the best of our knowledge, this study is the first to assess chemosensory alterations in 120 adults with obesity and COVID-19. We conducted a secondary analysis of cross-sectional 121 survey data collected between April 7th and November 4th, 2020 using the Global Consortium 122 for Chemosensory Research (GCCR) core questionnaire. This crowdsourced survey collected 123 data from community-dwelling individuals via social and traditional media as well as the GCCR 124 website. It was also presented to clinicians to relay to their patients. This survey, currently 125 deployed in 32 languages, used binary-response and categorical questions, as well as visual 126 analog scales to measure self-reported chemosensory ability and other symptoms in adults with 127 ongoing or recent respiratory illnesses (22). We also collected self-reported data on the presence 128 of pre-existing diseases, including our condition of interest, obesity, as well as other COVID-19 129 symptoms All participants included in the study were: 1) ≥ 18 years old, 2) had a (suspected) 130 respiratory illness within the past two weeks, 3) had onset of respiratory illness after January 1, 131 2020, 4) reported COVID-19 diagnosis via laboratory test (viral PCR or antigen test). 132 Respondents who did not report having any illness or symptoms within the last two weeks, who 133 had multiple responses, or who responded "Don't know" or "Other" when asked about their 134 diagnosis of COVID-19, were excluded from the analyses. To investigate the recovery of 135 chemosensory functions, only participants who reported the date of onset of respiratory illness 136 symptoms were included. The original study was approved by the Office of Research Protections 137 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10. 1101/2021 of The Pennsylvania State University (STUDY00014904). The hypotheses and analyses in this 138 manuscript were pre-registered at https://osf.io/xf25v (26) and the research compendium 139 including data files and analysis scripts is available at https://osf.io/rbcty/. Specifically, our 140 analyses aimed to describe chemosensory perception and related symptomatology during the 141 COVID-19 illness (Aim 1) and post-vs pre-COVID-19 diagnosis (Aim 2), in participants with 142 self-reported obesity vs without obesity. We predicted lower ratings for smell, taste, and 143 chemesthesis, and more severe COVID-19 symptoms in participants with obesity, compared to 144 without obesity. We also speculated smaller differences in ratings for smell, taste, and 145 chemesthesis perception post-vs pre-COVID diagnosis in participants with self-reported 146 obesity. Post-COVID-19 chemosensory recovery (Aim 3) was also tested, hypothesizing that 147 participants will have lower ratings for smell, taste, and chemesthesis post-COVID-19 recovery 148 in participants with self-reported obesity vs without obesity. Additionally, we assessed COVID-149 19 severity as measured based on the sum of reported symptoms (Aim 4), and the ability of smell 150 ratings to predict COVID-19 diagnosis (Aim 5), in participants with self-reported obesity vs 151 without obesity. 152 A departure from the pre-registered analyses is the inclusion of age as a factor in all 153 analyses, following differences in age we observed between groups. We also report the 154 unregistered analysis of pre-illness ratings, an important addition given the previously reported 155 decreased sensitivity for participants with obesity compared to those without obesity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10.1101/2021.02.28.21252536 doi: medRxiv preprint final analysis. A positive COVID-19 diagnosis (C19+) was determined using the self-reported 161 data from COVID-19 lab test or clinical exam outcome. All C19+ patients were further 162 categorized into having obesity if they reported it as one of the pre-existing disease conditions in 163 the questionnaire. C19+ patients who did not report having any medical condition or did not 164 answer this question on pre-existing disease conditions were categorized as controls without 165 obesity. We also included a control group of participants without obesity. See Cognizant of possible null effects in all our analyses, we opted to implement a Bayesian 180 approach, which allows us to estimate the strength of the evidence supporting the null 181 hypothesis. To test via a between-participant sequential Bayes factor design whether a difference 182 between groups was present (H1) or absent (H0), we conducted Bayesian linear regressions with 183 for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10.1101/2021.02.28.21252536 doi: medRxiv preprint the lmBF function from the BayesFactor package (28). We used the default Cauchy prior on the 184 effect sizes under the H1 as the scale parameter spread, which was set at its default value of r = 185 sqrt (2)/2. To test for a difference in age between groups, we used the following full model: Age 186 ~ COVID diagnosis + Obesity Age + COVID diagnosis x Obesity. Additive models (no 187 interaction) and main effect models were also computed and compared to determine the model 188 that best explained the data pattern, aka the model comparison with the most extreme Bayes 189 Factor. Please refer to Supplementary Table 1 for the inference rules, which follows the 190 classification scheme proposed by Lee and Wagenmakers (29) and adjusted from (30). To 191 interpret the strength and the direction of the effects identified, we have additionally sampled 192 from the models' posterior distributions (iterations = 1e4). To test for gender differences between 193 groups, we calculated probability tables of women and men in each of the COVID-19 and 194 obesity groups and tested for distribution differences with Pearson's chi-square tests with the R 195 base function "prop.test". We used an alpha of 0.05 to determine significance. 196 197

Self-reported Chemosensory perception analyses 198
For chemosensory perception analyses, we also conducted Bayesian linear regressions 199 with the lmBF function. The full model included the following terms: Dependent variable ~ 200 COVID diagnosis + Obesity Age + COVID diagnosis x Obesity. Additive models (no 201 interaction) and main effect models were also computed and compared to determine the model 202 that best explained the data pattern. Age was included in all models to factor in significant 203 associations between age and obesity. We used "before illness", "during illness", "change due to 204 illness" ("before illness" minus "during illness") and "recovery" ("after illness" minus "during 205 illness") separately as dependent variables. 206 for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10. 1101/2021 Other illness symptomatology analyses 208 To assess whether participants with obesity experience more and/or different symptoms 209 from those without obesity, we summed all symptoms that participants reported (each symptom 210 that was reported was assigned a value of 1). We then conducted Bayesian linear regressions 211 with the lmBF function as above with summed symptoms as the dependent variable (as above in 212 the chemosensory analyses). We operationalized disease duration as the number of days since 213 onset of the illness and used "days since onset" as the dependent variable in Bayesian linear 214 regression (models as above). For the subset of COVID-19 patients only, we calculated 215 probability tables for the likelihood of experiencing a given symptom for the participants with 216 and without obesity and tested for distribution differences with chi-square tests (details as above 217 under demographics). We used an alpha of 0.05 to determine significance. summarizes the tradeoff between sensitivity (fraction of correctly identified C19+ cases in the 226 sample with obesity and without obesity) and specificity (fraction of correctly identified C19-227 cases in the sample with obesity and without obesity) as the threshold value for the predictor is 228 varied. We used symptoms (binary), number of symptoms, chemosensory ratings during illness, 229 for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10.1101/2021.02.28.21252536 doi: medRxiv preprint ratings because those best showed evidence for the effects of illness and were also the most 231 predictive symptom in a previous study with the same questionnaire (9). Moreover, this question 232 (rather than pre-illness ratings or change in ratings) is best suited for being asked when making 233 an inventory of symptoms in a clinical setting. 234

COVID-19 illness as those without obesity 279
To further understand changes in chemosensory perception with COVID-19 diagnosis 280 and obesity condition, we looked at the data from participants who reported recovery from the 281 illness (Figure 4, Supplementary Table 6). Recovery was reported by 3970 participants 282 (n=3431 C19+ and n= 539 C19-), which is approximately 68% of our sample. Our Bayesian 283 linear models suggest that the ratings for post-recovery chemosensory perception (smell BF 10 = 284 7.79e-02 ± 0.02%; taste BF 10 = 6.44e-02 ± 0.02%; chemesthesis BF 10 = 2.18e-01 ± 0.01%) did 285 not differ in C19+ and C19-diagnosis. Of note, some smell/taste/chemosensory symptoms 286 remain post-recovery from the illness in C19+ and C19-groups. We found no differences in 287 smell (BF 10 = 6.58e-02 ± 0.02%), taste (BF 10 = 1.07e-01 ± 0.02%), and chemesthetic perception 288 (BF 10 = 9.96e-02 ± 0.11%) by self-reported obesity. Nasal obstruction did not seem to be 289 affected by either COVID-19 diagnosis (BF 10 = 5.33e-02 ± 0.03%) or obesity status (BF 10 290 =8.16e-02 ± 0.02%) post-recovery from the illness. 291 292 Participants with obesity report more symptoms overall and more frequently report 293 respiratory and gastrointestinal (GI) symptoms. 294 Based on the evidence from existing clinical and epidemiological studies, one of our 295 goals was to assess whether individuals with self-reported obesity overall have greater 296 symptomatic manifestation with C19+ diagnosis than those participants without obesity. To test 297 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10.1101/2021.02.28.21252536 doi: medRxiv preprint our hypothesis, we used Bayesian linear regression and compared the sum of the symptoms 298 reported by participants in these samples versus samples without obesity ( Figure 5A, 299 Supplementary Table 8). As predicted, among those with C19+, there is decisive evidence that 300 participants with obesity report a larger number of symptoms than participants without obesity 301 (BF 10 = 1.02e04 ±0%; average N of symptoms = with obesity: 8.22; without obesity: 7.42). A 302 similar effect is observed among participants with C19-(average N of symptoms = with obesity: 303 8; without obesity: 7.33 BF 10 = 9.91e03 ±0%). Among those with C19+, disease duration is 304 longer in those with obesity (BF 10 = 1.02e04 ±0%; average days since onset), while in C19-such 305 a difference is not observed (BF 10 = 1.21e00 ±0% (Figure 5B, Supplementary Table 7). 306 Looking at the specific symptoms (Figure 5C), smell and taste symptoms are equally reported 307 by participants with obesity and participants without obesity with a diagnosis of  Further, participants with self-reported obesity reported greater frequency in loss of appetite, 309 diarrhea, and nausea, along with shortness of breath, cough (dry or with mucus), and chest 310 tightness. 311 312 A classifier trained on participants without obesity accurately predicts C19+ diagnosis in 313 participants with obesity 314 Based on the self-reports on symptoms, combined with the chemosensory and nasal 315 obstruction ratings, we assessed the accuracy with which we could predict a C19+ diagnosis 316 ( Figure 6) in OB-. We then tested the model to predict the accuracy of discrimination of C19+ in 317 participants with obesity. Our results indicate that we can predict the C19+ diagnosis with 63% 318 accuracy, which indicates a moderately good estimate. Variables included in this analysis are 319 reported in Supplementary Table 8. 320 for use under a CC0 license.
This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

DISCUSSION 322
Since the beginning of the COVID-19 pandemic, reports of olfactory and gustatory 323 dysfunctions in COVID-19 patients continue to grow. To our knowledge, this study is the first to 324 describe and compare the chemosensory perception and related symptomatology in COVID-19 325 patients who self-reported to have obesity vs. no obesity. Independent of the obesity status, the 326 subjective ratings of smell, taste, and chemesthesis declined with COVID-19 illness. Examining 327 the recovery patterns, we found that participants with obesity show similar recovery from 328 COVID-19 related loss of smell, taste, and chemesthesis as those without obesity. Although we 329 do not know the severity of each symptom, those with obesity reported a greater frequency of 330 respiratory and GI symptoms and more symptoms overall. Finally, we found that a model of all 331 symptoms combined that was trained on patients without obesity is, can predict the C19+ 332 diagnosis with 63% of accuracy in participants with obesity. Furthermore, this smell loss was not 333 related to self-reported nasal obstruction, commonly observed in other upper respiratory 334 infections (33,34). Together, these results confirm and add to previous reports that COVID-19 335 largely impacts chemosensory function; however, obesity does not mask self-reported 336 chemosensory loss in those with a COVID diagnosis. 337 Smell and taste disturbances are a typical consequence of nasal inflammation due to an 338 upper respiratory tract viral infection (35,36); however, an acute loss of taste and smell emerged 339 rapidly as a critical neurological manifestation of a positive COVID-19 diagnosis (37). Our 340 current findings are similar to prior reports that showed that approximately 90% of the 341 participants reported a loss of smell. Furthermore, nearly 80% of the participants reported a loss 342 of taste, and 46% had a reduction of chemesthesis (detection of chemicals that induce tingling 343 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10. 1101/2021 and burning sensations such as the burning of chili peppers), indicating that the chemosensory 344 impairment is not restricted to smell (9,22). While most cold viruses cause nasal congestion and 345 individuals experience a reduction in the sense of smell, our results showed that nasal congestion 346 was not associated with smell loss. This finding is consistent with other reports (38-40) where 347 individuals with COVID-19 do not report clinically significant nasal congestion or rhinorrhea, 348 suggesting that other mechanisms may play a role in COVID-19 associated smell loss (37). 349 In addition to being a risk factor for COVID-19 viral infection, excessive body weight is 350 also implicated in chemosensory decline. Adipose tissue in obesity is "pro-inflammatory", 351 causing a surge in levels of IL-6 and C-reactive protein and enhancing the expression of 352 cytokines and adipokines (41). Interestingly, in diseases where these circulating inflammatory 353 factors are high, smell and taste dysfunction are prevalent (42,43). In particular, acute induction 354 of systemic inflammation has been shown to shorten the lifespan of adult taste bud cells (18). 355 Similarly, enhanced expression of inflammatory markers is shown to reduce olfactory sensory 356 neurons, in mice fed a high-fat diet to induce obesity (44). Thus, obesity-related inflammation 357 may affect chemosensory function. A major concern with this pre-existing gustatory and 358 olfactory sensory deficiency due to obesity is that obesity may mask the viral-induced 359 diminished taste and smell self-reported experiences. Interestingly, our analysis showed that 360 COVID-19 related chemosensory-related changes were comparable between C19+ participants 361 with obesity and without obesity suggesting that obesity does not have an effect on the loss of 362 chemosensory perception with COVID diagnosis. These findings need to be taken with caution, 363 especially when considering severe cases, which are more common in patients with obesity. For 364 example, if a patient is in critical condition, they cannot pay attention to their chemosensory 365 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 3, 2021. ; https://doi.org/10.1101/2021.02.28.21252536 doi: medRxiv preprint alterations, and chemosensory perception will likely not be tested or self-reported. This does not 366 mean that the chemosensory perception is not affected. 367 In terms of chemosensory recovery, we found no differences between participants with 368 obesity compared to those without obesity. While none of the studies to date have compared the 369 recovery rates between C19+ participants with obesity vs no obesity, our overall recovery rate of 370 65% is comparable to our previous analysis (9) but slightly lower than other studies (45,46). 371 There are residuals smell/taste/chemosensory symptoms reported post-recovery from the illness 372 in C19+ and C19-groups. In particular, quantitative studies using psychophysical methods have 373 shown that nearly 25% of people continue to report chemosensory problems when evaluated 30 -374 60 days after the onset of . This insufficient recovery rate may significantly 375 increase the number of patients with chemosensory disturbances, ultimately influencing eating 376 behaviors (47), quality of life (48,49), and psychological health (50,51) in the general population. 377 But most importantly, it may significantly impact participants with obesity who have an added 378 burden of lower chemosensory acuity due to excess fat mass (44,52). Thus, it is imperative to 379 prepare healthcare workers to detect and treat chemosensory disorders in this high-risk 380 population. 381 As we hypothesized, non-chemosensory symptoms were more severe in C19+ 382 participants with obesity than in participants without obesity. Specifically, participants with 383 obesity reported a greater frequency of respiratory and GI symptoms. In general, it is known that 384 obesity is associated with GI symptoms disturbances, such as upper abdominal pain, nausea, 385 vomiting, retching, and gastritis. GI symptoms are accompanied by inflammation or alterations 386 of intestinal permeability (53-56). However, it also emerged that COVID-19 patients 387 experienced several GI symptoms such as diarrhea (24.2%), anorexia (17.9%), and nausea 388 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.