An Explanation for Reports of Increased Prevalence of Olfactory Dysfunction With Omicron: Asymptomatic Infections

Abstract

The prevalence of olfactory dysfunction (OD) in people infected with the Omicron variant is substantially reduced compared with previous variants. However, 4 recent studies reported a greatly increased prevalence of OD with Omicron. We provide a likely explanation for these outlier studies and reveal a major methodological flaw. When the proportion of asymptomatic infections is large, studies on the prevalence of OD will examine and report predominantly on nonrepresentative cohorts, those with symptomatic subjects, thereby artificially inflating the prevalence of OD by up to 10-fold. Estimation of the true OD prevalence requires representative cohorts that include relevant fractions of asymptomatic cases.

The prevalence of olfactory dysfunction (OD) in the coronavirus disease 2019 (COVID-19) pandemic has fluctuated with the different variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1, 2]. Nearly all studies conducted to date have reported a decreased prevalence of OD in people infected with the Omicron variant, compared with infection with the Alpha or Delta variants [2]. Curiously, 4 recent studies from China reported the opposite—that Omicron causes a greatly increased prevalence of OD [3–6]. Is it possible that Omicron causes much less OD than previous variants in most of the world, but increases OD prevalence 10-fold exclusively in parts of China?

Here we explored plausible explanations for the seemingly discrepant reports. Since half of the outlier studies originated from Shanghai, China, we asked whether the reason for the high OD prevalence may be related to the exceptionally high proportion of asymptomatic cases in this region [7–9]. One conceivable hypothesis is that nonrepresentative cohorts—excluding the large fraction of asymptomatic cases—may concentrate the OD cases among the small number of symptomatic cases. Thus, the increased OD prevalence may be erroneous, caused by neglect of the large number of asymptomatic cases. Some previous studies concluded that asymptomatic cases among people infected with Omicron are 3-fold higher than with previous variants [10], but others [11] have reported the proportion of asymptomatic cases with Omicron to be similar to that with previous SARS-CoV-2 variants [12]. The proportion of asymptomatic cases appears to depend on multiple parameters, including not only variant type but also the age of the population [13]. Meta-analyses on the proportion of asymptomatic cases infected with Omicron are based on small numbers of studies and relatively small cohorts [10, 11], allowing only tentative conclusions.

We sought to provide support for our hypothesis that an inadvertently inflated OD prevalence may be caused by a large proportion of asymptomatic infections. We verified in a meta-regression analysis that the global OD prevalence has decreased with Omicron, and we geographically mapped the OD prevalence studies. We compared in a meta-regression analysis the proportion of asymptomatic infections, and we geographically mapped such studies for Omicron compared with pre-Omicron variants. We focused on studies of adults or mostly adults, because the proportions of asymptomatic infections and OD prevalence are known to differ in children and adolescents compared with adults [13, 14].

METHODS

For our systematic review and meta-analysis of OD with different variants of SARS-CoV-2, we made use of our own previous meta-analyses [1, 2], and we updated the previous data with additional studies, especially from world regions not previously covered. For the systematic review and meta-analysis of the proportion of asymptomatic infections with previous variants, we relied on earlier reviews [10–13] and updated the data as described below.

Search Strategy and Inclusion Criteria

Previous systematic reviews [1, 2] that estimate the prevalence of OD were updated by searching the iCite National Institutes of Health COVID portal (https://icite.od.nih.gov/COVID19/search/) with the keywords “Omicron” and “smell.” The studies reporting asymptomatic proportions were compiled from previous systematic reviews, for pre-Omicron variants [12, 13] as well as Omicron [10, 11], and updated by searching the National Institutes of Health COVID portal using the key words “Omicron” and “asymptomatic.” For data on OD and asymptomatic proportions, studies specifically on children were excluded, to avoid a major confounding variable. Only studies with >100 patients in the cohort were considered, so they could be geographically mapped. When the same data were published twice, only one data set (earlier publication date) was analyzed. For the studies included in our analysis, updated through 21 June 2023, see the lists (Supplementary Tables 1–4) in the Supplementary Materials. Data were extracted by using predesigned tables, including date of publication, first author name, country, geographic region, cohort size, number of cases, and the percentage calculated from the number of cases per cohort.

Statistical Analyses

Pooled analyses were performed for OD prevalence and proportion of asymptomatic infections, as described elsewhere [2]. The heterogeneity among studies was evaluated by means of the Cochran Q test and the I² index. Random-effects models were used to conservatively diminish the heterogeneity between the studies. A continuity correction of 0.5 was applied to studies with zero cells. Meta-regression analyses tested the difference in pooled prevalence or proportion between periods of pre-Omicron variants and Omicron dominance. The significance level was set to .05. All meta-analyses were performed using the Stata SE 16.0 software (StataCorp).

RESULTS

We verified that the global OD prevalence indeed decreased with Omicron. Our pre-Omicron data set had a pooled OD prevalence of 42.3% (95% confidence interval: 38.5%–46.0%) (Supplementary Figure 1), while the pooled OD prevalence with Omicron, including the 4 outlier studies, was 9.6% (7.4%–11.8%) (Supplementary Figure 2). This difference was significant (P < .001). The global distribution of the OD studies illustrates the decrease of OD prevalence with Omicron in a world map that compares the OD prevalence between different variants (Figure 1A and 1B). These maps are based on data from 168 studies with 101 553 adults (or mostly adults), depicting the pre-Omicron cohort sizes by the size of the circles, and the OD prevalence by the color gradient. This map was compared with 74 studies, reporting on 834 846 adults (or mostly adults) infected with the Omicron variant. The comparison confirms a substantial reduction, previously calculated to be a 4–10-fold lower prevalence of OD with Omicron [2]. Figure 1 also shows that the prevalence of OD in Western countries was, and still is, higher than in most of Asia (Supplementary Tables 1 and 2 and Supplementary Figures 1 and 2), but Omicron's OD is reduced by approximately the same amount in both regions [2]. Regarding the outlier studies from China, we note that they are globally unique in reporting increased OD prevalence with Omicron.

Omicron can cause a large proportion of asymptomatic cases. The difference compared with pre-Omicron variants was estimated by previous meta-analyses to be 3-fold higher with Omicron than with Delta [10], but this conclusion was based on a small number of studies and cohorts. Other studies reported proportions of asymptomatic cases to be comparable for pre-Omicron variants and Omicron [11, 12]. Because some of this heterogeneity could be due to the mixing of pediatric and adult studies (and it is known that children have a higher proportion of asymptomatic cases), we analyzed only studies with adult or mostly adult cohorts. Our meta-analysis of a larger number of studies (pre-Omicron, 119 studies; Omicron, 44 studies) and cohorts (pre-Omicron, 378 640 patients; Omicron, 1 551 450 patients) found a nearly identical pooled proportion of asymptomatic cases: 32.8% pre-Omicron (95% confidence interval: 25.7%–39.8%) versus 32.5% (25.8%–39.1%) with Omicron (Supplementary Figures 3 and 4), a difference that is not statistically significant (P = .96). This is also illustrated in the global map of the distribution of studies reporting the proportion of asymptomatic cases (Figure 2A and 2B, Supplementary Tables 3 and 4, and Supplementary Figures. 3 and 4). The largest proportion of asymptomatic infections is concentrated in certain regions of China, especially in Shanghai (Figure 2B). This suggests (but does not prove) that the outlier OD prevalence may be linked to the exceptionally high proportion of asymptomatic cases.

DISCUSSION

Data obtained by our systematic reviews and meta-analyses support the hypothesis that methodological issues are largely responsible for the outlier discrepant reports. When the proportion of asymptomatic cases is large, this has important consequences for the calculation of the OD prevalence. Such prevalence studies typically determine the number of OD cases among the number of symptomatic patients—asymptomatic cases are rarely included. This means that the true denominator (symptomatic plus asymptomatic cases) is much larger than the number of symptomatic cases alone, resulting in a substantially overestimated prevalence of OD. For example, when the prevalence of OD is calculated in a cohort of only symptomatic cases (even though 90% of Omicron cases in the general population are asymptomatic), then 40 cases with OD among 100 symptomatic cases would be reported as a 40% prevalence of OD. The true prevalence is only 4% (40 of 1000), because for every 100 cases of symptomatic infection, there are 900 cases of asymptomatic infection that should be included in the denominator but often are not. When the prevalence of OD was higher (as with previous variants), the proportion of asymptomatic cases among infected cases would have changed OD prevalence much less (and therefore may have gone unnoticed), because the denominator mismatch would have had a much smaller effect on the prevalence calculations.

Consistent with our hypothesis, the proportion of asymptomatic cases among Omicron-infected people has been reported to be very high, between 90% and 95% in Shanghai, China, in the spring of 2022 [7–9], and this was not due to inclusion of presymptomatic cases (only 3.8% converted from asymptomatic to symptomatic [7]). The unusually high proportion of asymptomatic cases in Shanghai was confirmed by multiple large-scale studies [7–9] (Supplementary Table 4 and Supplementary Figure 4). The high proportion of asymptomatic infections in the population of Shanghai likely explains why 2 of the 4 OD studies report an increased OD prevalence with Omicron [3, 4] (Figure 1B). The apparent outlier OD prevalence in these Shanghai studies is due to a skewed denominator, because the cohort of symptomatic patients was not representative of the entire infected population. Indeed, the first outlier study [3] examined only symptomatic patients, hospitalized with mild symptoms, and the second [4] examined a cohort with at least 79% symptomatic patients from makeshift hospitals (the exact numbers were not disclosed), which does not match the 5%–10% proportion of symptomatic infections in the general population of Shanghai in the spring of 2022 [7–9].

The proportion of asymptomatic cases in populations infected with Omicron is highly variable according to the I² values [10, 11], with several populations exceeding 90% asymptomatic infections—such high proportions of asymptomatic cases have rarely been reported in large cohorts of adults infected with pre-Omicron variants (Supplementary Figure 3). Heterogeneity between studies may be caused by a variety of factors, including differences in the definition of asymptomatic status, verification of asymptomatic status at different time points, differences in age, in vaccination status, extent of previous infections and antibody status, society-specific countermeasures of transmission (compulsory test regimens), different subvariants, and yet-to-be-identified parameters [7, 10–13]. Previous (pre-Omicron) OD studies likely also had some bias due to different amounts of asymptomatic infections, but with a truly higher OD prevalence and with a lower proportion of asymptomatic cases than the extremes in Shanghai, the deviations from the genuine OD prevalence would be expected to be much smaller.

A second, but conceptually related, reason for a potentially misleading reporting of OD prevalence is that internet surveys may have recruitment bias, because persons with symptoms tend to be more motivated to respond, resulting in an “enrichment” of patients with OD in the cohort [2, 5]. Therefore, the prevalence data from such surveys can be biased and may not be representative of all those infected with Omicron, as previously elaborated [2]. The third and fourth of the outlier studies [5, 6] are examples where recruitment bias likely contributed to the seemingly 10-fold increased OD prevalence compared with other studies (Figure 1B). These surveys’ heavy focus on chemosensory dysfunction leaves the impression to prospective participants that this was the sole study goal [5, 6]. A nonrepresentative cohort is further evidenced by the cohort's 79% female proportion [6]. An increased OD due to a lack of previous exposure to SARS-CoV-2 is unlikely, because the large majority of studies from China report decreased OD with Omicron [2].

To determine Omicron's true prevalence of OD and taste dysfunction, it is essential to either recruit representative fractions of symptomatic and asymptomatic people infected with Omicron or to clarify that only a subset of patients, those with symptoms, were recruited and examined. Only studies with representative cohorts report the true OD prevalence. The prevalence of OD (and of any chemosensory dysfunction) is decreased with the Omicron variant, for reasons that likely involve the less effective endosomal virus entry route instead of the route using host cell surface-membrane fusion [15]. Estimating the true numbers of people with chemosensory dysfunction is important for healthcare planning and for arranging appropriate treatments. Grossly incorrect and misleading estimates of numbers of OD cases hinder the timely and effective utilization of scarce resources.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Disclaimer. The sponsor had no role in the study design, the collection, analysis, or interpretation of data, the writing of the report, or the decision to submit the report for publication.

Financial support. This work was supported by the National Institute of General Medical Sciences, National Institutes of Health (grant GM103554 to C. S. v. B.).

References

Author notes