Prevalence of Olfactory Dysfunction with the Omicron Variant of SARS-CoV-2: A Systematic Review and Meta-analysis

Abstract

The omicron variant is thought to cause less olfactory dysfunction than previous variants of SARS-CoV-2, but the reported prevalence differs greatly between populations and studies. Our systematic review and meta-analysis provide information about regional differences in prevalence as well as an estimate of the global prevalence of olfactory dysfunction based on 62 studies reporting on 626,035 patients infected with the omicron variant. Our estimate of the omicron-induced prevalence of olfactory dysfunction in populations of European ancestry is 11.7%, while it is significantly lower in all other populations, ranging between 1.9% and 4.9%. When ethnic differences and population sizes are taken into account, the global prevalence of omicron-induced olfactory dysfunction in adults is estimated at 3.7%. Omicron’s effect on olfaction is twofold to tenfold lower than that of the alpha or delta variant, according to previous meta-analyses and our analysis of studies that directly compared prevalence of olfactory dysfunction between omicron and previous variants. The profile of prevalence differences between ethnicities mirrors the results of a recent genome-wide association study that implicated a gene locus encoding an odorant-metabolizing enzyme, UDP glycosyltransferase, to be linked to the extent of COVID-related loss of smell. Our analysis is consistent with the hypothesis that this enzyme contributes to the observed population differences.

Figure 1.

Flowchart illustrating the Literature Search, Systematic Review and Meta-analysis according to the PRISMA guidelines. PubMed, the NIH COVID portal, and Google Scholar were searched sytematically; other sources were found in references of articles and in news media. The literature was last updated on January 10, 2023. The reasons for exclusion of full-text articles were, in decreasing frequency: no prevalence data; no data for omicron; long (not acute) COVID; children only; review; case report.

Figure 2.

World map showing the location of cohorts included in the systematic review and the prevalence of olfactory dysfunction due to the omicron variant. The size of the circles represents the size of the cohort as indicated in blue; the color gradient indicates the prevalence range as shown on the right side. Note that populations of European ancestry have larger prevalences than populations of non-European ancestry.

Figure 3.

Forest plot of the 62 studies reporting the prevalence of olfactory dysfunction due to the omicron variant. The confidence intervals (CI) and the weight of each study are indicated on the right. The pooled overall global prevalence is 6.6% according to the meta-analysis, but this does not take into account ethnic differences and population sizes as explained in Fig. 4. The size of the light grey box is proportional to the study weight. The study weights are obtained based on the DerSimonian-Laird method. CI, confidence interval; DL, DerSimonian-Laird method; I², I-squared index.

Figure 4.

Forest plots of the prevalence of olfactory dysfunction due to omicron differentiated by regions/ethnic populations according to the meta-analysis. Prevalences are 3.1% in populations in Africa (CI = 1.6%–4.6%), 11.7% in people of European ancestry (Western countries, CI = 10.3%–13.1%), 2.8% in South Asia (CI = 0%–6.3%), 1.9% in East Asia (CI = 1.2%–2.7%), 2.2% in the Middle East (CI = 0.3%–4.1%), and 4.9% in Latinos/Hispanics (CI = 3.5%–6.2%). The size of the light grey box is proportional to the study weight. The study weights are obtained based on the DerSimonian-Laird method. CI, confidence interval; DL, DerSimonian-Laird method; I², I-squared index.

Figure 5.

The prevalence of olfactory dysfunction (OD) due to omicron is reduced by 2-fold to 10-fold compared to the previous SARS-CoV-2 variants, varying by ethnic population and region. (A) The reduction of OD due to omicron in direct comparisons within similar populations and regions during the predominance of mostly the delta variant (for specifics of the comparator variants, see Table 2). The size of the light grey box is proportional to the study weight. CI, confidence interval; DL, DerSimonian-Laird method; I², I-squared index, RR, risk ratio. (B) The bar graph summarizes the reduction in prevalence of OD for the direct comparisons from panel A (n=32), and also two indirect comparisons with pooled estimates from previous meta-analyses, von Bartheld et al., 2020 [99], and Mutiawati et al., 2021 [100]. The percent reduction ranges between 2-fold and 10-fold.

Figure 6.

Funnel plots for the Western studies (A) and the studies for all other ethnicities (B) reporting omicron-induced olfactory dysfunction. These are scatterplots of prevalences against their standard errors. The vertical solid line is the estimated effect size; the dotted lines are the corresponding pseudo 95% confidence intervals (CIs). They provide insight into the spread of observed effect sizes. The majority of studies are randomly scattered within the CI region, indicating absence of publication bias for Western studies (A, p=0.591) and for all other ethnicities (B, p=0.703).

Figure 7.

The bar graph shows the differences between ethnicities for omicron’s prevalence of olfactory dysfunction (blue bars), compared with the frequency of the risk allele for olfactory dysfunction in the UGT2A1 locus according to Shelton et al., 2022 [13] (orange bars). The blue bars show the pooled estimate for the hyposmia prevalence of the same ethnicities according to our systematic review. Meta regression shows that there is a positive association between the two parameters (p<0.001). The prevalence of olfactory dysfunction is higher in those ethnic populations that have a higher frequency of the risk allele. The two exponential trend lines show this similarity.

Figure 8.

The pooled prevalences of olfactory dysfunction due to omicron (A) and due to other variants (B) in survey-type studies and traditional-design studies for populations of European ancestry. While the pooled estimates are higher in the survey-type studies than in the traditional-design studies,14.2% vs.10.9% for omicron studies (A), and 45.4% vs 36.6% for previous variants (B), meta-regression showed no heterogeneity between the two study types. Confidence intervals for omicron studies (A) were 9.7%–18.7% for survey-type studies, and 9.3%–12.6% for traditional design studies. Confidence intervals for previous variant studies (B) were 22.1%–68.8% for survey-type studies, and 28.4%–44.8% for traditional design studies.