What to propose to treat post-viral non-conductive smell disorder in COVID-19?-systematic review with meta-analysis

Klinger Vagner Teixeira da Costa (  klingercostamcz@gmail.com ) Cesmac University Center https://orcid.org/0000-0003-1425-1348 Aline Tenório Lins Carnaúba Cesmac University Center https://orcid.org/0000-0003-4100-6866 Katianne Wanderley Rocha Cesmac University Center https://orcid.org/0000-0002-0577-5740 Kelly Cristina Lyra de Andrade State University of Health Sciences of Alagoas (UNCISAL) https://orcid.org/0000-0002-2618-4958 Pedro de L. Menezes State University of Health Sciences of Alagoas (UNCISAL) https://orcid.org/0000-0003-1999-5055


Introduction
Non-conductive smell disorder (NCSD) is characterized by not being associated with nasal obstruction. Its etiology is often di cult to be identi ed, and therefore, indicating the best treatment is a challenge. There are several etiologies for acquired NCSD, and the most common are post-viral, post-traumatic and neurological [1]. The post-viral NCSD is the most common one [2,3], with variable duration, and 70% of cases occurring by rhinovirus, in uenza and parain uenza viruses, respiratory syncytial virus, adenovirus and severe acute respiratory syndrome virus 2 (SARS-CoV-2) (4). Severity can vary from mild to severe, and it can be permanent [2].
In the last few months, an increasing number of cases of olfactory disorders (ODs) associated with the coronavirus disease 2019 (COVID-19) caused by SARS-CoV-2 has been observed [5]. Currently, ODs are recognized as part of the COVID-19 clinical feature, with the onset before the general symptoms/signs of COVID-19, and most often not associated with nasal obstruction or rhinorrhea as in other viral infections, that is, being clinically characterized as a NCSD [6][7][8][9]. The recovery of smell in COVID-19 occurs, in most cases, in the rst two weeks after resolution of the disease, while 26.4% remain with ODs, and it is still unknown whether long-term recovery will occur [9].
Unfortunately, no effective standard therapy for NCSD of any etiology has been established, even at NCSD in COVID-19. Several drugs have been proposed, for example: corticosteroids (systemic, topical), zinc, theophylline, alpha-lipoic acid, caroverine, vitamin B and others [10][11][12]; however, the effectiveness of most of these treatments remains uncertain. On the other hand, it has been experimentally shown in animals and in clinical studies that olfactory training with different odors may improve olfactory function [13][14][15]. Improvement in the knowledge on a speci c topic can be supported by systematic reviews (SR) with meta-analyzes. They identify, analyze and synthesize the best scienti c evidence for best clinical practices, especially in the face of new clinical scenarios such as the current COVID-19 pandemic and the signi cant increase in ODs cases, in which randomized controlled trials and precise recommendations for their treatment are still unavailable.
The objective of this SR with meta-analysis is to identify and analyze the best scienti c evidence on the effectiveness of the treatments available for NCSD to propose the best therapeutic options for olfactory function recovery in the face of the increasing demand from patients with olfactory sequelae caused by COVID-19.

Methods
This systematic review (SR) with meta-analysis aimed to answer the following question: "Which drug therapies and/or olfactory therapies are most effective for treating patients with non-conductive smell disorder?", according to PICO strategy for research. This SR is in accordance with the items of the Preferred Reporting Items for Systematic Reviews and Meta-Analyzes Statement (PRISMA) [16]. The protocol was registered on June 8, 2020 in the International Prospective Register of Systematic Reviews -Prospero database (https://www.crd.york.ac.uk/PROSPERO/) under registration number CRD42020189452

Search strategy
The strategies aimed at a complete search, including descriptors (Medical Subject Headings -Mesh) and free terms in order to identify relevant articles within the proposed question. The following search strategy was developed: (anosmia OR loss smell OR smell loss OR olfactory dysfunction OR smell disorder OR olfactory disorders) AND (treatment OR corticoids OR acid, alpha lipoic OR alpha lipoic acid OR lipoic acid OR zinc chelated pentosan polysulfate OR zinc OR chelated zinc OR melatonin OR sodium citrate OR omega-3 OR omega-3 fatty acid OR -3 fatty acid OR vitamin D OR vitamins OR olfactory training OR olfactory therapy); this strategy was adapted to the databases. No language or date restrictions were applied. The complete search strategies are shown in the supplementary material (Appendix 1).
Between June 10 th and 13 th , 2020, the following electronic databases were searched: Medline via PubMed, Cochrane Clinical Trials, ScienceDirect, Lilacs, Scopus, Google Schoolar and Web of Science, as well as the gray literature OpenGrey.eu, DissOnline, The New York Academy of Medicine and Research Gate. To avoid citation bias, there was no manual search for the included articles and specialists in the area were not contacted [17].
After the searches were performed, the references from each database were exported to Mendeley® (https://www.mendeley.com/) to identify duplicate articles, promote greater reliability in the selection and proceed to the eligibility criteria step.

Eligibility criteria
The following inclusion criteria were considered: (1) clinical trials with individuals diagnosed with non-conductive smell disorders; (2) quanti cation of the degree of olfactory capacity through a validated method before and after the intervention. The exclusion criteria were considered: (1) studies on neurological olfactory disorders; (2) studies in which the intervention was not related to the treatment of the olfactory disorder; (3) duplicate publications.

Data extraction
The article search was performed by two evaluators. The titles and abstracts of the retrieved articles were independently assessed by two researchers who were not blind to the authors or the journal titles. The full versions of potentially eligible articles were retrieved for further assessment, carried out in three phases: 1. Identi cation and reading of titles in different electronic databases. Articles that clearly did not meet any of the inclusion or exclusion criteria were excluded.
2. Reading of abstracts of the studies selected in the rst phase. Likewise, articles that clearly did not meet any of the inclusion or exclusion criteria were excluded.
3. The studies which were not excluded in these rst two phases were read in full for the selection of articles to be included in this SR with meta-analysis.
The main data of each article was fully collected and inserted into a database structured in the Microsoft O ce Excel 2010® software.
There were no divergences in the evaluators' analysis. A standard form for data storage was created based on the model adopted by Cochrane [18]. For better presentation of the results, we decided to consider the following variables of the selected articles: Title, author, sample size (N), age, olfactory evaluation method, NCSD causes and treatment. The outcomes sought in the studies were both the treatments proposed for olfactory disorders and the evidence of their effectiveness.
Quality assessment of studies The quality of the studies was assessed according to the recommendations on the Cochrane Collaboration handbook [19]. Two investigators independently assessed the quality of the studies in the following categories: random sequence generation, allocation concealment, blinding of participants and professionals, blinding of outcome assessment, incomplete outcomes, selective outcome reporting and other sources of bias.
After these assessments, the selected studies were submitted to a statistical analysis to verify the possibility of a meta-analysis. This analysis combines and summarizes the results of several studies, thus increasing the accuracy and power of the evidence of the results.
When necessary, the characteristics of the study considered as potential sources of heterogeneity were included in a subgroup analysis.
In addition, in case of high heterogeneity, the studies were removed, one by one, to identify which one was the source of high heterogeneity. The pooled results were meta-analyzed using a xed effects model if the studies were homogeneous or of low heterogeneity; a random effect model was used if the studies showed moderate or high heterogeneity. An α <0.05 value was considered statistically signi cant. When adequate data for analysis was impossible, Cochrane's recommendations were followed.
All analyzes were performed using RevMan software (version 5.3, The Nordic Cochrane Center, The Cochrane Collaboration, Copenhagen, Denmark, 2014).

Included studies
In the rst phase of this SR with meta-analysis, 5,743 articles were found in seven databases, but none in the gray literature. After excluding 304 duplicate studies, 5,439 were selected for reading titles and abstracts. 5,420 were excluded according to the established selection criteria (1,829 studies with patients with ODs associated with nasal obstruction or neurological causes; 1,220 due to missing information on treatments; 1,190 without olfactory data; 512 literature reviews; 668 experimental studies; and one Cohort), and 19 articles were selected for full reading. After the reading, six articles were excluded because their samples were composed of patients with obstructive nasal disease (table 1). However, of the 13 studies included in the qualitative analysis (table 2), ve studies were unique in their treatment [10,[21][22][23][24] and four did not have enough data for quantitative analysis (meta-analysis) [25][26][27][28]. Thus, four studies were selected for the meta-analysis involving 336 patients [29][30][31][32] (table 3). The entire article selection process is described in gure 1, which shows the PRISMA ow diagram for inclusion.

Risk of bias assessment
Of the 13 included articles, three studies speci ed the blinding methods of participants and professionals, as well as the blinding of outcome assessors providing enough information to allow judgment [22,24,27].
Most studies presented a risk of uncertain bias, mainly due to the impossibility of judging by the following criteria: random sequence generation and allocation concealment, denoting an important selection bias; and many studies presented risk of uncertain bias for the item blinding of participants/professionals and for the item blinding of outcome assessments, denoting measurement bias.
The study by Reden et al. (2017), reports how Vitamin A was administered in the groups as a limitation [24]; the study by Poletti et al. (2016), states that possible differences in solubility may interfere in the effect of the molecular weight of odors [30]. The quality analysis of the included articles is shown in table 4.

Data analysis
Regarding the general characteristics of the articles included in the qualitative analysis, seven studies were conducted in Germany and six in other countries (Table 1). Outcome measures were determined by symptom scores and objective methods of olfactory testingwith the Sni n`Sticks test as the most common test used in the studies. Ten studies compared the mean TDI score ("T" -threshold, "D"discrimination, "I" -identi cation) before and after drug treatment in NCSD (total of six drugs) [10,[21][22][23][24][25][26][27][28], while four compared the mean TDI score before and after olfactory training in NCSD [29][30][31][32] (table 3). Of the studies with drug treatment, ve were the only ones for the drug they proposed to analyze (oral caroverine, oral vitamin A, intranasal insulin, intravenous pentoxifylline and oral lipoic acid) [10, [21][22][23][24] and four were on intranasal sodium citrate [26][27][28]; however, studies with sodium citrate did not provide su cient pre-and posttreatment data for meta-analysis.
Drug treatments for NCSD.

1-Caroverine
In a randomized controlled study, Quint et al. investigated the use of quinoxaline derivative caroverine to treat various sensorineural ODs, including post-viral NCSD [21]. A total of 77 patients with NCSD were included in the study, 51 received oral caroverine for 4 weeks (120 mg/day) and 26 controls for 1 month. Olfactory sensitivity was assessed before and after treatment, and patients were blinded. The tests included the assessment of the odor threshold of n-butanol (BTT) and odor identi cation (Sni n'Sticks test). Caroverine improved the thresholds (p <0.005) and odor identi cation (p <0.042) in patients with anosmia, and signi cantly improved the ability to identify odors (p <0.041) in patients with hyposmia.
The results showed that 13 of the 35 patients with anosmia in the treatment group improved the thresholds in BTT (6 responders had post-viral NCSD). Transitory nausea was reported by two patients and transitory headache by one patient. No severe adverse effects were reported.

2-Vitamin A
In a double-blind, controlled and randomized clinical trial with oral vitamin A in 52 patients with post-viral and post-traumatic NCSD, 26 participants received oral vitamin A (10,000 IU/day) and 26 received placebo for three months [24]. The olfactory function was assessed through the Sni n'Sticks test in which the patients were blinded, and through a standardized questionnaire. Forty-four percent of patients reported smell recovery, and 29% showed signi cant improvement in olfactory function. However, there was no signi cant difference between the results of patients who received oral vitamin A or placebo.

3-Insulin
In a double-blind, controlled and randomized clinical trial with intranasal insulin in 38 patients with NCSD, 19 participants received intranasal insulin (40 IU), and 19 received placebo (saline) twice a week for four weeks [22]. Olfactory function was assessed through the Connecticut Chemosensory Clinical Research Center (CCCRC) test at onset and four weeks after treatment.
The CCCRC test score in the intervention group was signi cantly higher in comparison with the placebo group (p = 0.01). Intranasal insulin appears to be effective in the treatment of NCSD.

4-Alpha-lipoic acid
A prospective, uncontrolled clinical trial investigated 23 patients with post-viral NCSD treated with oral alpha-lipoic acid. The administered regimen included alpha-lipoic acid (600 mg/day), for approximately 4.5 months [10]; there was no control group. The olfactory function was assessed through the Sni n'Sticks test and the patients were blinded. Of the treated patients, 26% showed a moderate improvement in the TDI score (1.5≤TDI <5.5 points), and 35% showed a notable improvement (TDI≥5.5). The recovery of olfactory function appeared to be more evidenced (t = 2.56 [p = 0. 018]) in younger patients than in patients over 60 years (the age of 60 was the average of the group). Oral alpha-lipoic acid appears to be effective in the treatment of post-viral NCSD.

5-Pentoxifylline
In a longitudinal study with pentoxifylline to treat diseases of the inner ear, 19 patients received intravenous pentoxifylline (400mg/day) [23]. Olfactory function was assessed through the Sni n'Sticks test before and after drug treatment. After pentoxifylline administration, patients demonstrated a decrease in the odor threshold towards lower odor concentrations (p = 0.01) and was increased in younger patients than in the elderly adults.

6-Sodium citrate
In the rst randomized clinical trial by Whitcroft et al. (2016), with 57 patients with NCSD (post-viral, post-traumatic or idiopathic), one nostril was treated with sodium citrate (3.5g /140ml) and the other nostril with 0.9% sodium chloride [25]. The Sni n 'Sticks test was performed separately in each nostril. The patients were blinded to the side of the sodium citrate application. Compared with placebo, there was a signi cant improvement in the identi cation scores of participants with post-viral NCSD after treatment with sodium citrate.
In the second clinical trial by Whitcroft et al. (2016) with 49 patients with post-viral NCSD, one nostril was treated with sodium citrate (3.5g/140ml) and the other nostril with 0.9% sodium chloride [28]. The Sni n 'Sticks test was performed separately in each nostril. The patients were blinded to the side of the sodium citrate application. There was a statistically signi cant improvement in olfactory thresholds after treatment with sodium citrate compared with placebo in all patients (mean improvement of 0.87 ± 2.68 points, p = 0.04) and in the analysis of subgroups in those with hyposmia (mean improvement of 1.15 ± 2.37 points, p = 0.02).
In a randomized clinical trial with 31 patients with NCSD (post-viral, post-traumatic or idiopathic), each patient's olfaction was assessed twice [26]. On the rst day, olfaction was tested before and after intranasal administration of 0.9% sodium chloride in each nostril; on the third day, the same patients' olfaction was tested before and after intranasal administration of sodium citrate (3.5g/140ml). The olfactory function was assessed through the Sni n'Sticks test. Thirty patients (96.8%) improved their scores in less than an hour after using sodium citrate.
In a randomized, double-blind clinical trial with 55 patients with NCSD (post-viral, post-traumatic or idiopathic), 31 participants received intranasal sodium citrate (3.5g/140ml), and 24 received 0.9% sodium chloride [27]. The olfactory thresholds were assessed through a test using phenylethyl alcohol (roses), 1-butanol (pear), acetic acid (vinegar) and eucalyptol (menthol), previously validated by the same authors. A signi cant effect was observed in the intervention group for phenylethyl alcohol and 1-butanol and eucalyptol when compared to control group (p <0.05); 32% of patients treated with sodium citrate showed an improvement in sensitivity to some of the odors.
Thus, intranasal sodium citrate appears to be effective in the treatment of NCSD, including post-viral cases, with a temporary effect lasting between 30 and 60 minutes.

Effect of olfactory training on NCSD
On the topic olfactory training, four articles and 336 patients were selected; the articles studied olfactory training with four odors: phenyl ethanol (rose), menthol (mint), citronellal (lemon) and eugenol (clove), twice a day (morning and night), instructing the patient to smell each substance for 10 seconds, and 10 seconds of interval between them.
The research carried out the meta-analysis of the four articles that performed olfactory training for at least three months; the improvement in olfaction (increase in mean TDI score) after olfactory training (SMD = 5.99; 95% con dence interval (CI) [4.68, 7.31], p <0.00001) was statistically higher than in pre-training ( g. 2). Considering the causes of NCSD in these articles, all studies had patients with sequelae after viral infection, and in one of them, the sample was also composed of patients with post-traumatic sequelae and idiopathic causes [32]; therefore, a subgroup meta-analysis was carried out with the three studies whose sample was composed with only post-viral NCSD cases and with the data referring to the group with post-viral NCSD from the study with mixed sample; this subgroup meta-analysis also showed an improvement in olfaction (increased mean TDI score) after olfactory training (SMD = 5.18; 95% con dence interval (CI) [4.12, 6.23], p <0.00001) was statistically higher than in the pre-training ( g.3); the data from the used olfactory training and the results can be seen in table 2. The olfactory training improved the mean TDI score in all studies, and especially in patients with post-viral NCSD.

Discussion
The current SR with meta-analysis was performed to identify and analyze the best scienti c evidence on the effectiveness of treatments available for NCSD in order to propose the best therapeutic options to recover olfactory function due to the increasing number of patients with olfactory sequelae caused by COVID-19. The results of this review can be classi ed into two lines of treatment: one with administration of drugs and the other with olfactory training.
Regarding the drug treatment, nine articles and 401 patients were selected; the articles studied six substances, two of them with intranasal use (insulin and sodium citrate); three with oral via (caroverine, alpha-lipoic acid and vitamin A) and intravenous pentoxifylline.
However, only sodium citrate was found in more than one article (total of four), while the other substances were studied by unique studies. Regarding the four studies on sodium citrate, the data were insu cient to perform a meta-analysis and, therefore, remained only in the qualitative analysis.
Oral caroverin (120mg/day) for one month was effective in improving the threshold and odor identi cation in post-viral NCSD; however, patients reported transitory nausea during treatment. Caroverine may modulate the activity of mitral cells in the olfactory bulb through its antagonism at N-methyl-D-aspartate (NMDA) receptors improving the detection of olfactory-induced activation [33]. Other mechanisms may include competitive antagonism at both alpha-amino-3-hydroxy-5-methylisoxazole-4-proprionic acid (AMPA) receptor sites and, at higher concentrations, at the NMDA receptor's glycine site [34]. As excitation of granule cells through mitral-tufted cells is strongly NMDA receptor-dependent, blocking of NMDA receptors might also reduce g-aminobutyric acid negative feed-back onto mitral and tufted cells, which in turn might increase net excitation in the olfactory bulb [33].
Intranasal insulin (40 IU) for four weeks improved olfactory function in patients with NCSD, but studies with groups of NCSD speci cally post-viral are missing. The mechanism of intranasal insulin is quite clear, as the olfactory bulb is the largest place for the transformation of intranasal insulin to the brain [35].
Alpha-lipoic acid (600mg/day) for about 4.5 months was shown to be effective, mainly in younger patients with post-viral NCSD. It stimulates the expression of nerve growth factor, substance P, and neu-ropeptide Y [36][37][38]; it also has potent antioxidative effects and neuroprotective capabilities indicating that lipoic acid is suited to treat neural damage involving free radicals [22].
Pentoxifylline (400mg/day) has been shown to improve the olfactory threshold, but there are no further studies on its use speci cally in post-viral NCSD. Because pentoxifylline is an unespeci c phosphodiesterases (PDE) inhibitor, it can result in increased intracellular concentration of cAMP; elevated cAMP level leads to increased response of olfactory sensory neurons to chemical stimuli [39].
Intranasal sodium citrate (3.5g/140ml, pH 7.4, osmolarity 298) was the most studied substance, and in all studies there was an improvement in olfactory function in patients with post-viral NCSD; however, it has a transitory effect which lasts about 30-60 minutes.
Sodium citrate acts as sequestrant, binding to free Ca ++ ions in the nasal mucus and therefore diminishing mucosal Ca ++ ; it leads to ampli ed receptor responses, improving olfactory function [26].
Regarding olfactory training, all studies had samples composed of patients with post-viral NCSD; however, the study by Hummel et al.
(2009), did not specify the number of patients in the sample with post-viral, post-traumatic or idiopathic NCSD [32]. The number of women was higher than that of men and the average age of the groups was over 50 years. NCSD degrees were not mentioned in all studies, but all groups submitted to olfactory training intervention, for at least three months, showed improvement in olfactory function.
The results of the meta-analyzes showed that olfactory training is effective in the rehabilitation of olfactory function in cases of postviral NCSD.
Electrophysiological studies have shown that repeated exposure to odors in humans can produce increased electrical potential records from the olfactory epithelium [40]. However, the positive effect of olfactory training may not only be related to peripheral factors, but may possibly be attributed to central changes [14]. A study shows that there are representations for odor categories in the olfactory bulb [41]; thus, the use of different odor categories in a training scheme can induce more prolonged activation of olfactory function.
The olfactory function has the ability to recover when subjected to repeated short-term exposure to odors due to an increase in the growth of neurons in the olfactory receptor and an increased expression of receptors in response to odor exposure [42]. Therefore, olfactory training should produce a general increase in olfactory function.
Therefore, in drug treatment, intranasal sodium citrate can be used to minimize the olfactory limitation in NCSD, especially during meals, as a medication with topical effect and with transitory duration. Thus, sodium citrate and olfactory training appear as therapeutic proposals to recover olfactory function in post-viral NCSD, such as in COVID-19, requiring additional clinical trials to con rm the evaluated results and establish treatment protocols.

Limitations
This systematic review with meta-analysis has some limitations. Although electronic searches were performed in multiple databases, some relevant studies may have been overlooked in our search strategy. Although all of the included studies are clinical trials, many showed uncertain bias, mainly due to the possibility of judging the criteria: random sequence generation and allocation concealment, blinding of participants and outcome assessors, promoting a lesser probability to determine a cause-effect relationship between events.
The studies were also conducted across different countries, which can lead to different results in olfactory tests due to sociocultural differences. Also, the pooled data were limited by a small number of included studies and small sample size and, therefore, more prone to bias. Overall, these limitations should be considered to indicate a treatment for NCSD due to COVID-19.

Conclusion
This study demonstrated that there is scienti c evidence about the e cacy of drug treatment and olfactory training in recovery of olfactory function in post-viral non-conductive smell disorder.    Figure 1 Identi cation and selection ow chart. Olfactory training in patients with NCSD: difference in the mean of TDI score before and after training.

Figure 3
Olfactory training in patients with post-viral NCSD: difference in the mean of TDI score before and after training.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Appendix1.docx