Olfactory recovery following infection with COVID-19: A systematic review

Olfactory loss has been identified as one of the common symptoms related to COVID-19 infection. Although olfactory loss is recognized, our understanding of both the extent of loss and time to olfactory recovery following infection is less well known. Similarly, knowledge of potential impactful patient factors and therapies that influence olfactory recovery is desirable but is not overtly clear in the literature. Our systematic review sought to fill this knowledge gap. We included studies that: involved either an observational or an interventional design that reported data on patients with olfactory dysfunction due to Reverse Transcription Polymerase Chain Reaction (RT-PCR) diagnosed COVID-19 infection; and reported data regarding olfactory recovery measured by an objective olfactory test, Likert scale and/or visual analog scale (VAS). The study methods were determined a priori and registered in PROSPERO (Registration Number CRD42020204354). An information specialist searched Medline, Embase, LitCovid and the Cochrane Register of Controlled Trials up to March 2021, and two reviewers were involved in all aspects of study selection and data collection. After screening 2788 citations, a total of 44 studies of assorted observational designs were included. Patients had undergone objective COVID-19 testing, and most were adult patients with mild to moderate COVID-19. Olfactory recovery was found to occur as early as 7 days, with most patients recovering olfaction within 30 days. Few studies included prolonged follow-up to 6 months or longer duration. Poor olfaction at initial presentation was associated with poor recovery rates. Only a small number of studies assessed olfactory retraining and steroid therapy. Additional trials are underway.


Introduction
The spread of Coronavirus  infection was announced by the World Health Organization to be a pandemic on March 11 th 2020 [1]. Since the pandemic began, more than 178

Search strategy
Searches to identify relevant studies for this review were conducted by an experienced medical information specialist (RS) in consultation with the study authors. Literature searching was conducted in the following databases: Ovid MEDLINE, Embase, the Cochrane Central Register of Controlled Trials (CENTRAL) and LitCovid. The MEDLINE search strategies were peer reviewed by a senior information specialist (BS) using the PRESS checklist prior to extraction [12]. Systematic reviews on this topic were also retrieved from MEDLINE and their reference lists were scanned for additional potentially eligible studies to be screened. The searches were initially conducted on July 21 st , 2020, were last updated in May 2021. A copy of the search strategies is provided in the Review Supplement (S1 Text).

Study design.
Studies involving either an observational or interventional design were of interest. No restrictions related to follow-up duration were employed. We excluded all review articles, case reports, letters, editorials, commentaries, and abstracts.
Outcomes. The primary outcome of interest was the extent of olfaction recovery following COVID-19 infection, using objective or subjective measures. Therefore, a baseline measurement of olfactory loss at time of infection or at study intake was compared to subsequent measurements at follow up intervals. Olfactory loss could be reported as a continuous or categorical measurement (anosmia, hyposmia, normosmia). Our secondary outcomes of interest included time to olfactory recovery, patient prognostic factors (for occurrence and resolution of olfactory dysfunction), and interventional therapies used for olfactory recovery.
Inclusion was limited to studies published in either the English, French, or Spanish languages.

Process of study selection
The search yield from the literature search was imported into COVIDENCE software (Veritas Health Innovation, Version v2632 03f7ed40) [13]. Duplicate citations were removed prior to the start of screening. Selection of papers was conducted in two phases involving two independent reviewers (AJ and AL). During the first phase, each reviewer screened all titles/abstracts to make judgements as to relevance to the objectives and any conflict in the decision to include for further assessment was resolved by consensus discussion or consultation of a third party (SK). If the content of the abstract was unclear, the article was selected for full text review. In phase 2, full text articles of the citations from phase 1 deemed potentially eligible were screened using our predetermined inclusion and exclusion criteria. Any conflicts in the final decision to include a study were again resolved through consensus discussion or consultation of a third party (SK). The final process of study selection has been summarized using a PRISMA flow diagram (Fig 1).

Process of data extraction
Analogous to our study selection process, two reviewers (AJ, AL) extracted data independently from the set of included studies using standardized data extraction form implemented in COVIDENCE software [13]. Discrepancies were settled by consensus discussion amongst the two reviewers. The data were then exported to Microsoft Excel (Version 16.49; Microsoft Corporation, Seattle, Washington, USA) for further preparation and to allow for inspection by team members to consider between study heterogeneity and study findings.
The following data were extracted from each study:

Risk of bias assessment
The risk of bias of the included studies was evaluated using relevant tools developed by the Joanna Briggs Institute (JBI) critical appraisal checklist [14]. The JBI assessment tool has separate assessments for each study design included in our systemic review. Case series as well as cross-sectional, cohort, and case-control studies were evaluated using the appropriate domains as set per the JBI assessment tools. Two reviewers (AJ, AL) independently evaluated the risk of bias of eligible studies and any disagreements were resolved by consensus discussion. Results from the risk of bias assessments were summarized descriptively to present key strengths and weaknesses of the included studies. The detailed findings from risk of bias assessments are presented in the supplemental files to this review.

Approach to data synthesis
Due to considerable heterogeneity in study/patient characteristics and endpoints, no metaanalyses of outcomes were performed. A descriptive approach to summary of outcomes of interest was instead prepared. Grouping of studies has been based upon the type of olfactory assessment tool used, looking separately at studies that examined objective versus subjective measures. Study data have been organized and presented in table format which maps the outcomes reported by individual studies. The aggregate data of each study characteristic were reported, as well as testing instrument, initial olfaction scores, follow up time points, and prognostic factors as well as any therapeutics implemented. Combination of graphical and numerical displays of findings are presented for outcomes where appropriate.

Study selection process
Our search identified a total of 2,788 citations for review following removal of duplicates. Amongst them, 2,546 were found to be irrelevant, and the remaining 241 were assessed for eligibility in full text. Overall, 46 publications describing 44 unique studies met our inclusion criteria, and 200 were excluded for reasons detailed in the flow diagram in Fig 1.

Study population
All studies consisted of a patient population of asymptomatic or mild to moderate COVID-19 symptoms. Not all studies reported the scale used to classify the severity of COVID-19 but, in general, asymptomatic patients and those with mild symptoms did not require medical intervention or hospitalization (other than for isolation reasons). Patients with moderate symptoms required medical intervention but not admission to an Intensive Care Unit. The hospitalized patients included in all studies were of mild to moderate symptoms, none of which required invasive ventilation in ICU. COVID-19 was confirmed by RT-PCR test in all but six studies that either used RT-PCR or serology as a diagnostic tool of COVID-19 [25,48,51,54,58,60]. Although most of the patients recruited within studies were not hospitalized due to COVID-19 infection [17, 21-26, 28, 30-32, 35, 36, 41-43, 45, 47-49, 51, 54, 56, 60], some did include patients hospitalized either for monitoring purposes or quarantine [18,19,33,52,55,[57][58][59]. Twelve studies included a mixture of hospitalized and non hospitalized patients [15, 19, 20, 27, [24,51], during active phases of symptomatology [47], from the time of the swab [35,36], at initial home/clinic visit [27,30], at the time of admission to hospital [29,52], during the recovery period [17,22], after recovery/discharge  [49], or at other specified times [41,54,55]. The timing and number of follow-up evaluations also varied amongst studies. The majority of studies performed one follow-up olfactory re-test following their first initial testing, and thirteen studies performed more than one follow-up evaluation [20, 23, 24, 26, 28, 34-36, 43, 51, 52, 57, 60].

Findings from risk of bias appraisals
The evaluations using the JBI critical appraisal checklists showed several limitations in the available evidence. Risk of bias was considered high for the studies that performed the evaluation of olfactory function at a single point in time given the risk for recall bias [19, 38-40, 42, 48, 50, 56]. The only randomized controlled trial included in this review did not blind participants or clinicians to the assignment of participants to the use of therapy, nasal steroids, therefore the risk of bias for this study was considered high [49]. In one pilot study, participants were not randomly assigned to the study arm, which has the potential to introduce selection bias [25]. S1 Table shows the results of the evaluations for each study.

Extent and time to recovery
Reporting of initial baseline anosmia and hyposmia was inconsistent amongst included studies. In those studies that performed objective testing, less than 40% of the population reported normosmia at initial evaluation compared to those who performed subjective testing, where normosmia could range from 0% to 90% at initial evaluation. Studies with follow up data regarding the extent of recovery of olfactory function and the time to recovery were analysed. A total of 4 studies that used objective olfactory measures completed follow-up data at 1 month [15,24,27,28]; rates of full olfactory recovery in these studies ranged from 44.3% to 94.6% (median 72.6%). Seven studies completed follow-up data for up to 60 days [16, 20-23, 59, 60], and rates of full recovery in these studies ranged from 0% to 79.5% (median 73.3%).

Protective and risk factors for olfactory dysfunction and resolution
Approaches to explore associations between olfactory resolution and other patient characteristics varied between studies. This included both univariate and multivariable regression analyses, subgroup analyses, and correlation studies. Amongst the most commonly studied factors were age, gender, medical co-morbidities, and initial severity of smell loss. Studies evaluating the effects of age determined it was not a factor that contributed to poor recovery rate [16, 18-20, 25, 29, 46, 60]. One study showed age was inversely associated with improvement of olfactory function [15]. A positive association between age and the duration of hyposmia/anosmia was reported by one study [49] another study reported that persistent olfactory dysfunction is associated with age (� 50 years) [26]. When the effect of gender on olfactory recovery was assessed, the majority of studies did not show any significant relationship between recovery and gender [15,16,26,29,32,38,46,60], while one study found that females represent more than 70% of those who completely recovered [34] and one study showed female gender showed slower olfactory recovery rates [53]. Many studies had specific population eligibility criteria, thus excluding patients with an a priori diagnosis of neurodegenerative disease, asthma, craniofacial trauma, CRS, nasal polyposis disease. However, two studies identified that diabetes mellitus diagnosis amongst patients was a poor prognostic factor for recovery [34,49] while others did not identify medical co-morbidities including diabetes mellitus to be have any prognostic value [19,20,31,54]. An observational negative prognostic factor commonly identified was the initial degree of smell loss. Patients with more severe olfactory dysfunction at initial presentation had worse olfactory recovery outcomes compared to those with very mild to moderate hyposmia at initial presentation [18,20,30,54]. S2 Table presents a summary of studies that assessed the effects of different characteristics, as well as the nature of the association that was reported by the study authors.

Treatment intervention(s)
There was a paucity of studies in the literature that assessed interventions for the treatment of olfactory dysfunction following COVID-19 infection. A randomized controlled trial by Abdelalim et al 2021 [49] compared topical corticosteroid nasal spray (2 puffs, 100 mcg, mometasone furoate each nostril) once daily for 3 weeks in addition to olfactory training vs olfactory training only. This study included 100 patients who had recovered from COVID-19, demonstrated by 2 negative RT-PCR tests, but who had not recovered their olfactory function; Olfactory dysfunction (OD) was measured using a VAS. The study concluded that the use of nasal steroid spray showed no added benefit over olfactory training alone. A pilot study by LeBon et al 2021 [25] compared two groups; Group 1 received 32 mg of methylprednisone orally once daily combined with olfactory training twice daily for 10 weeks compared to olfactory training only. This study included 27 patients who had recovered from COVID-19 but showed persistent OD five weeks after the onset of symptoms. The Sniffin' Sticks test was used to measure OD. The outcome showed that combination therapy may be beneficial, although robust conclusions could not be made due to low recruitment of participants [25]. Vaira et al 2021 [23], conducted a case-control study in which patients were randomly assigned to systemic corticosteroids (prednisone starting with 1mg/kg day) tapering for 15 days as well as nasal irrigation with betamethasone, ambroxol and rinazine for 15 days or no treatment. This study included 18 patients who had recovered from COVID-19, demonstrated by 2 negative tests, and who still demonstrated anosmia or hyposmia 30 days after clinical onset measured using the CCCRC test. In this study, patients in the treatment group showed significantly higher improvement in the CCRC test at both the 20 and 40-day evaluations. We searched the Clini-calTrials.gov registry in March 2021 to identify ongoing clinical trials assessing the efficacy of treatment of olfactory dysfunction related to COVID-19. We identified 9 trials, (NCT04764981, NCT04710394, NCT04361474, NCT04657809, NCT04406584, NCT04495816, NCT04569825, NCT04789499, NCT04528329) and we have provided their details in S3 Table. These studies are evaluating the effects of therapies that include olfactory training, systemic and intranasal corticosteroids, insulin fast dissolving film, theophylline nasal irrigations and injections of platelet rich plasma into the olfactory cleft.

Discussion
In this study, the evidence regarding olfactory recovery following COVID-19 infection was reviewed. Our literature review identified 44 studies that met our inclusion criteria that used objective and/or subjective olfactory evaluation. We identified considerable heterogeneity between the studies in terms of study designs employed, psychophysical testing tools utilised, the timing of initial assessments, follow-up intervals, and recording of olfactory scores. Due to this high degree of between-study variability, we did not perform meta-analyses of outcome data, and results were presented using a descriptive approach.
Olfactory recovery most often occurred within the first two weeks from symptom onset with a rate of recovery at 1 month as high as 94.6%; at 6 months the rate of recovery was as high as 85.7%. We analyzed any treatment interventions as our secondary outcome. Only

PLOS ONE
three published papers assessed treatment for persistent OD due to COVID-19 [23,25,49]. The treatments included oral and nasal corticosteroids and olfactory training. The small numbers of studies evaluating treatments and the similarly small numbers of patients included in these studies limit the ability to make firm conclusions. However, it has been demonstrated that olfactory training is associated with clinical improvement of olfactory dysfunction following upper respiratory tract viral infection [61,62] and therefore it is a reasonable treatment alternative. There are several ongoing studies (NCT04764981, NCT04710394, NCT04361474, NCT04657809, NCT04406584, NCT04495816, NCT04569825, NCT04789499, NCT04528329) evaluating olfactory training and once completed, they will provide insight on whether this treatment is also effective in olfactory dysfunction due to COVID-19. The use of corticosteroids in post viral olfactory dysfunctions is not widely recommended [62][63][64] and their use in OD due to COVID-19 needs further study. We identified that patients with initial poor olfactory scores had slow rates of recovery, however the majority of studies did not observe medical co-morbidities, demographics, or general symptoms to be a pertinent factor in olfactory recovery.

Strengths, limitations and implications for clinical practice
More than a year has passed since the WHO declared the COVID-19 pandemic. Many countries have experienced multiple waves and more recently the surfacing of numerous widespread virus variants. Concerns over olfactory dysfunction have become a common complaint of many patients surviving COVID-19 and a source of referral to clinicians. The data presented in this review can be used by healthcare professionals to counsel patients regarding the duration and expected recovery from OD. The major strength of our study lies in the fact that all the studies included in this review used RT-PCR to confirm COVID-19. We included studies that used objective and/or subjective measures for olfactory dysfunction. It is well known that olfactory dysfunction may be underreported when using subjective methods and thus the use of objective tests is preferred. However, the information provided by the studies using subjective methods may be used by clinicians in settings where access to objective methods is limited. There was considerably a high level of heterogeneity across studies that restricted our ability to carry out a pooled statistical analysis. Some studies, especially those from early in the pandemic, performed a retrospective assessment of olfaction which is subject to recall bias. Although we updated our literature search twice, the speed at which new studies related to COVID-19 are being published makes it difficult to include all available evidence in this review. Updates to this review will be needed in the future. We expect that newer studies will include randomized controlled trials evaluating therapies for OD with larger numbers of participants and longitudinal studies that will have longer follow up times.

Conclusion
Our study identified several important and clinically relevant factors related to olfactory dysfunction following COVID-19 infection. We noted that the majority of studies showed olfactory recovery as early as the first two weeks, although some studies with more prolonged follow-up showed a small population of patients can have residual olfactory loss even after 6 months. Poor olfaction at initial presentation was associated with poor recovery rates, and factors such as age, gender, or medical co-morbidities did not play a major role in olfactory recovery. Treatment interventions included steroid therapy (topical/systemic) and/or olfactory retraining, but no robust conclusions were possible. Future research in this area should include large longitudinal studies that follow the course of the olfactory dysfunction beyond 6 months and well designed RCTs that are powered to determine effectiveness of olfactory training and other therapies for the treatment of persistent olfactory dysfunction due to COVID-19.