1. ABOUT THE DATASET ------------ Title: Output from questionnaire relating to NCT04868435 study entitled 'Triggers for post-viral parosmia' Creator(s): Jane Parker and Lisa Methven Organisation(s): University of Reading Rights-holder(s): University of Reading Publication Year: 2022 Description: Data from a six-part questionnaire looking at triggers for parosmia (olfactory distortions). The 6 parts comprised questions on 1) demographics, 2) loss of smell, 3) onset of parosmia 4) loss of taste, 5) responses for 14 set trigger foods and 6) other triggers and comments. 954 people started the questionnaire. After removal of trial runs and non-completers, data from the remaining 727 are presented here. Cite as: Parker JK, Methven L, (2022): Output from questionnaire relating to NCT04868435 study entitled 'Triggers for post-viral parosmia'. University of Reading. Dataset. https://doi.org/10.17864/1947.000367 Related publication: Pellegrino R, Smith BC, Gane S and Kelly CE (2022), Foods: Emerging Pattern of Parosmia Post COVID-19 and its Effect on Food Perception (March 2022 Under revision) Contact: j.k.parker@reading.ac.uk 2. TERMS OF USE ----------------- Copyright 2022 University of Reading. This dataset is licensed under a Creative Commons Attribution 4.0 International Licence: https://creativecommons.org/licenses/by/4.0/. 3. PROJECT AND FUNDING INFORMATION ------------ Title: Triggers for post-viral parosmia Dates: 19th June 2020-5th Sept 2021 Participating institutions: University of Reading, Monell Chemical Senses Center, School of Advanced Study, University of London, AbScent UK, Royal National Ear, Nose and Throat and Eastman Dental Hospitals, University College London Hospital Funding organisation: n/a Grant no.: n/a 4. CONTENTS ------------ File listing Output from Compusense questionnaire ReadMeParkerMethvenParosmiaQuestionnaire.txt ParkerMethvenQuestionnaire.docx ParkerMethvenParticipantInformationSheet.docx ParkerMethvenConsentForm.docx ParkerMethvenCompusenseOutput.xlsx 5. METHODS -------------------------- Ethics and Recruitment All subjects gave their informed consent for inclusion before they participated in the study. The investigations were carried out following the rules of the Declaration of Helsinki of 1975 and the protocol was approved by the School of Chemistry, Food and Pharmacy Research Ethics Committee of the University of Reading on the 10th of June 2020 (study number 29.2020). It was registered under the US Library of Medicine as trial NCT04868435. This is a cross sectional study. Participants were recruited through ENT clinics, Facebook (AbScent Parosmia and Phantosmia Support group and personal accounts) and Twitter between 19th June 2020 and 5th September 2021. Volunteers aged 18 or over who were experiencing smell distortions, or for whom everyday things smelled different, odd, or disgusting were invited to participate in the fully anonymised survey. Entry into the study was dependent on participants completing a standard unlinked online consent form which then took them to the survey landing page. Participation in this online study was voluntary and respondents received no remuneration. The survey was carried out on Compusense (Ontario, Canada). The Questionnaire After completing the consent form, respondents completed the six-part survey. Demographic data (age, gender, country of residence, ethnic group, and smoking status) were collected in Section 1, whilst Section 2 asked questions about the speed, timings, and aetiology of the respondents' initial loss of smell (anosmia). They were asked when they lost their sense of smell (date), and the likely aetiology of their symptoms (COVID-19, other viral illness, accident (including head or brain injury), unexplained (or idiopathic), or other/do not know). Where the cause was COVID-19 or other viral infection, the speed of loss was reported as one of 4 categories: very suddenly before onset of other symptoms of infection, very suddenly during infection, very suddenly after infection, or gradually. In cases where it was attributed to COVID-19, further questions were asked about diagnosis (PCR, antibody test or no test) and severity. Section 3 asked about the onset of parosmia, whether it had been preceded by any partial or full recovery of sense of smell, (none, a few hints, any partial or full recovery) and whether the symptoms fluctuated significantly (four categories; no fluctuation, infrequent or minor fluctuations, significant daily random fluctuations, significant but generally get better during the day). In Section 4 respondents were asked to indicate whether they could taste salt and sugar (three categories; taste as normal, taste weaker, can't taste), whether they could detect heat in spices (yes/no), and whether they had experienced any metallic taste (yes/no) or burning sensations in their nose or throat (yes/no). The core survey (Section 5) concerned the respondents' perception of 14 foods (referred to as the 14 'set triggers') which had been pre-selected based on data from other informal studies. Onion, meat, coffee, onions, and eggs were selected on the basis of being among the most commonly reported trigger foods for parosmia. Chocolate, peanuts, bacon, fried foods, peppers, cucumber, and melon were selected because they were known but less frequent triggers, whereas butter, apple and rose were selected as examples of 'safe' foods and smells that were less likely to trigger parosmia. For each item, the respondents were asked to record whether they perceived the smell of that item as distorted (four categories; smells like it did before, smells distorted, I can't smell at all or I am not familiar with this food/smell so cannot answer). If either of the last two answers was selected, the survey skipped to the next item. Those who selected 'distorted' were asked to provide two to three words to describe the distortion. Next, respondents were asked to rate their hedonic assessment of the smell as pleasant (score =1), neither pleasant nor unpleasant (2), unpleasant (3) or so bad I want to gag/vomit/leave the room (4). An additional option had been provided in the questionnaire 'This food has always smelt unpleasant' to allow for those where there was no change in hedonic rating because the item had always been perceived as unpleasant. This option was used in 160 out of 6447 observations and these data were excluded on the grounds that there had been no change in hedonic valence due to parosmia, but we suspect that this answer may have been misinterpreted by many of the respondents. The ratio of the frequency of reports of negative valence (unpleasant and gag-inducing) to the sum of neutral and pleasant was calculated for each food. Lastly, respondents were asked to record the strength of the smell now in comparison to before smell loss: weaker than before (score = 1), same as before (2) or stronger than before (3) with the additional option to report that the intensity fluctuated (not associated with a score). Faecal odour had been highlighted in a previous publication based on social media as being less unpleasant and more tolerable for some, whilst for others there was a switch in hedonic valence from repulsive to pleasant. This was explored further by asking about the distortion of faecal odour on a three-category scale (same as before, distorted, or cannot smell), and about the hedonic quality on a two-category scale (no longer unpleasant, just as unpleasant as before). Section 6 involved a check-all-that-apply (CATA) question covering an additional 20 possible triggers selected to cover a wide range of food, drink and some environmental or personal care items, with the opportunity to add further triggers as free text. The final question gave the respondents the opportunity to add any further comments. Data Analysis Data relating to respondents' demographics, aetiology, onset, and recovery were expressed as total count (n) and proportion (%). To investigate associations between the different aetiologies and onset, partial recovery and frequency of fluctuation, contingency tables were prepared on the counts and analysed using Fisher's Exact test (p=0.05). To determine whether there were significant differences in taste loss between respondents that had suffered COVID-19 versus other viral infections, the count data were similarly analysed by Fisher's Exact test (α=0.05). The Kruskal-Wallis two-tailed test with multiple pairwise comparisons using Boniferroni correction was used to determine whether disgust was significantly different between the set triggers. The Kruskal-Wallis two-tailed test with multiple pairwise comparisons using Dunn's procedure was used to determine whether strength was significantly different between the set triggers (from Section 5). Kruskal-Wallis and Dunn's procedure were similarly used to determine whether there was a relationship between distortion and both hedonic valence and strength, and between strength and hedonic valence. Statistical significance was considered at the 5% level (p=0.05). Descriptions of distorted food items (including faecal odour) were cleaned, and words were spell-checked with Hunspell using a large English dictionary (https://cran.r-project.org/web/packages/hunspell/index.html). For the word clouds, single word or compound adjectives were extracted from the descriptions and qualifiers suggesting qualitative changes (weaker, stronger, faint etc.) were removed. Obvious synonyms were combined (e.g., gasoline/petrol, garbage/trash/bin, toxic/poisonous, poo/poop/faeces/feces/faecal/fecal odour, synthetic/artificial, cat-food/dogfood) and words with the same root were combined under one term (e.g., chemical/chemically, earth/earthy, burnt/burning rotten/rotting, but sick and sickly for example were deemed to relate to different smells). The frequencies of the words reported for each trigger were calculated and words where the frequency per item was never more than 1 were removed. This was carried out for the 14 set triggers as well as for the answers to the question on faecal odour. Words for each food were visually represented in word clouds with the size representing the frequency using ggwordcloud (Pennec, E., & Slowikowski, K. (2018), ggwordcloud: A Word Cloud Geom for ggplot2. (0.3.0) [Computer software]). Descriptions of distortions next underwent sentiment analysis, using the sentimentR package (Rinker, T. W. (2019), sentimentr: Calculate Text Polarity Sentiment (2.7.1) [R Package]) for each item and was averaged then compared using an ANOVA (with post-hoc analysis and Tukey's HSD). The 120 words were further split into descriptive words where there was a true description of aroma character (79), hedonic words where there was a clear valence attributed to the word (30) and the remaining words (11) (e.g., indescribable, different, funky, unusual). Principal component analysis (PCA) using covariance was carried out on the frequencies of the descriptive words, and on frequencies of the hedonic words. Manual counts of the items mentioned in the free text were performed in order to identify the most frequently reported triggers. All those items previously assessed in either the set triggers (14) or the CATA (20) were disregarded, as were complex dishes that contained a number of potential triggers (curry, falafel, pasta sauce, baked beans) and the focus was on simple ingredients and personal care, home care or environmental odours. These were counted in a word search, using the word or the root of the word where the word was commonly misspelt or had regional variations. However, each incidence was verified since the contextual significance in which these words were mentioned varied: many people chose to tell us which items were not distorted, or which items came back first, or some words were used as descriptors for others (e.g., coffee smells like bleach, so bleach was disregarded on that occasion).