Introduction

Olfactory dysfunction is the most specific symptom of acute COVID-19 [1–3] and our ability to smell can be heavily impacted by the disease [4]. In many patients, olfactory function is regained after the acute phase [5], yet a non-negligeable proportion of patients exhibits chronic dysfunction [6]. Estimations of the prevalence of olfactory dysfunction 6 months after COVID-19, the threshold for chronicity of olfactory dysfunction, vary widely [5,7–13]. A recent assessment of olfactory dysfunction based on subjective reports suggests that a staggering 61% still experience olfactory dysfunction two years after infection [6].

Despite the clear evidence that COVID-19 can result in long-term olfactory dysfunction, there are several caveats in current prevalence estimates that may account for the large differences in reported estimates. First, most of the estimations are based on self-reported olfactory dysfunction which is notoriously unreliable [14,15]. Second, COVID-19 status is seldom determined by biological assays but rather by self-reported status, which can bias the estimations. Third, the prevalence of olfactory dysfunction unrelated to COVID-19 is usually not considered. In fact, olfactory dysfunction in the general population was estimated to be around 20% before the pandemic [16–19], with leading causes being sinonasal diseases, traumatic brain injuries, and viral infections of the upper respiratory tract [20]. Thus, an accurate estimate of the prevalence of olfactory dysfunction after COVID-19 infection requires consideration of the prevalence of olfactory dysfunction in a sample where COVID-19 infection can be ruled out.

Olfactory dysfunction encompasses various impairments to the sense of smell, categorized into quantitative dysfunction, involving a reduction or loss of olfactory function (hyposmia or anosmia) [19], and qualitative dysfunction, which involves altered odor perception such as parosmia or phantosmia. Standardized assessments of parosmia and phantosmia can currently only be carried out using questionnaires [21] due to their subjective nature. While self-evaluation questionnaires are commonly used to assess quantitative dysfunction, their precision warrants caution in interpretation [14,15]. Therefore, when estimating the prevalence of olfactory dysfunction after COVID-19, a comprehensive approach should incorporate both questionnaires and well-validated gold standard psychophysical tests [22,23] for a more accurate assessment of objective and subjective aspects.

Healthcare workers may be particularly affected by long-term consequences of COVID-19. For example, mental health disorders were particularly high in this population both during and following COVID-19 [24,25]. In line with this, nearly half of infected healthcare workers reported reduced chemosensory abilities one year after COVID-19 [5], which is considerably higher than in the general population [26]. It is therefore particularly valuable to study the long-term effects of COVID-19 on olfactory function in this group. We designed this study to assess olfactory dysfunction in a sample of healthcare workers where continuous serum assessments of SARS-CoV-2 infection were available. Specifically, we included two groups of participants; one group that had undergone SARS-CoV-2 infection and a control group that had not, as confirmed by the continuous serum testing. Our objective was to establish the prevalence of olfactory dysfunction due to COVID-19 up to 15 months after infection by evaluating olfactory dysfunction with validated olfactory tests in both groups. We hypothesized that the group who had undergone SARS-CoV-2 infection would be afflicted by a significantly higher prevalence of olfactory dysfunction compared to the control group.

Materials and method

Results

First, we assessed quantitative olfactory dysfunction. To do so, we explored potential differences in subjective performance: here, participants from the COVID+ group (N = 98) evaluated their olfactory function as significantly worse, 6.9 (SD ±2.5) out of a total of 10 points compared to the COVID- group (N = 41), 8.9 points (SD ±1.2; W = 955; N = 139, p < .0001). In line with the subjective experience, the average Sniffin’ Sticks TDI score was also significantly lower in the COVID+ group according to a Wilcoxon rank sum test, 30.9 points (SD ±5.9), compared to the COVID- group, 34.0 points (SD ±3.4; W = 1416.5; N = 139, p = .006; Fig 1A). Interestingly, upon separate analysis of the TDI subscales, only discrimination (W = 1293; N = 139, p = .0008; Fig 1C) and identification (W = 1213.5; N = 139, p = .0002; Fig 1D) subscores were significantly lower in the COVID+ group (D = 11.66 ± 2.5, I = 12.11 ± 2.5) compared to the COVID- group (D = 13.19 ± 1.9, I = 13.68 ± 1.3). However, no statistically significant difference (W = 2107; N = 139, p = .65; Fig 1B) was observed for the threshold subscore between the two groups (COVID+ = 7.10 ± 2.4, COVID- = 7.11 ± 2.0). We then assessed the proportion of individuals with a clinically relevant quantitative olfactory dysfunction. Based on the TDI score, the frequency of quantitative olfactory dysfunction was significantly higher (X² (1, N = 139) = 6.28, p = .01) in the COVID+ group compared to the COVID- group. A total of 37% of the COVID+ group suffered quantitative olfactory dysfunction, with 4 individuals exhibiting anosmia (TDI score = 14.88 ± 2.0) and 32 individuals exhibiting hyposmia (TDI score = 26.11 ± 3.9). In the COVID- group, 20% (N = 8) showed quantitative olfactory dysfunction, all of whom had hyposmia (TDI score = 29.56 ± 0.8). The average TDI scores for individuals with normosmia were 34.38 (SD = 2.6) and 35.06 (SD = 2.8) for the COVID+ and the COVID- groups, respectively. Last, we investigated whether the degree of quantitative olfactory dysfunction was influenced by the time since COVID-19 infection and individuals’ age. A Spearman correlation test indicated that the TDI score of individuals from the COVID+ group was not significantly affected by the number of days since infection (N = 71, r = -.1, p = .43; see S1A Fig). However, TDI significantly declined with age in the COVID+ group (Spearman correlation test, r = -.28, p = .005, S1B Fig). It’s worth noting that in the COVID- group, a similar negative relationship between TDI and age was observed, although it did not reach statistical significance (r = -.27, p = .09, S1B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Quantitative olfactory dysfunctions for COVID-19 positive (COVID+, N = 98) and COVID-19 naïve (COVID-, N = 41) groups.

(A) Mean TDI scores for COVID+ and COVID- groups with error bars denoting standard deviation (SD) as well as values for each participant indicated by circles. Dashed lines indicate cut-off scores for the clinical diagnoses Hyposmia (threshold score: 30.75) and Anosmia (threshold score: 16.25). Higher TDI scores (possible range 1–48) indicate better olfactory performance. TDI subscale scores for the COVID+ and COVID- groups are presented with error bars denoting standard deviation (SD) for (B) Threshold, (C) Discrimination, and (D) Identification scores separately.

https://doi.org/10.1371/journal.pone.0306290.g001

Next, we assessed to what level individuals in the respective groups experienced qualitative olfactory dysfunction using the 4-item parosmia scale. The average score was significantly higher in the COVID+ group, 2.8 points (SD ±2.8), compared to the COVID- group, 0.7 points (SD ±1.2; Wilcoxon rank sum, W = 2595; N = 132, p < .0001; Fig 2A). Correspondingly, we found that significantly more individuals from the COVID+ group reported parosmia (48 of 94 participants, 51%) relative to individuals from the COVID- group (2 of 38 participants, 5%; X² (1, N = 132) = 22.2, p < .0001). To understand the degree of symptom severity, we also assessed the answers to the parosmia severity question in each group. As can be seen in Fig 2B, slightly more than half of the individuals in the COVID+ group who reported parosmia symptoms experienced only minor symptoms (None 49%, Mild, 28%, Medium 16%, Severe 7%). This means that about a quarter of all individuals in the COVID+ group experienced medium to severe parosmia symptoms 15 months after COVID-19.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Distribution of parosmia score.

A) Distribution of scores per patient group (N = 94), COVID-19 positive (COVID+) and COVID-19 naïve (COVID-, N = 38), on the Landis parosmia questionnaire. B) Percentage of individuals in each patient group, grouped by reported parosmia severity score. C) Proportion of COVID+ participants (N = 94) in each olfactory dysfunction classification group.

https://doi.org/10.1371/journal.pone.0306290.g002

Further, we followed up with 48 individuals from the COVID+ group with established parosmia 2.6 years after COVID-19 infection. Of the 41 individuals who responded, 23 (56% of sub-sample, 24% of total initial COVID+ sample) still experienced parosmia. In fact, 41% indicated that they experienced medium to severe parosmia symptoms (None 43.9%, Mild, 14.6%, Medium 31.7%, Severe 9.8%). This long-lasting qualitative olfactory dysfunction was also reflected by a high parosmia score (3.1 ± 2.3 points).

As to be expected, there was a considerable comorbidity between diagnoses. While 35% of the COVID+ group (total N = 94) experienced no olfactory dysfunction at all, 14% experienced anosmia/hyposmia but no parosmia, 29% experienced parosmia but no anosmia/hyposmia, and 22% experienced both anosmia/hyposmia and parosmia. In other words, a total of 65% of COVID+ participants experienced some form of olfactory dysfunction on average 15 months after COVID-19 (Fig 2C). This frequency was lower in the COVID- group (N = 38) where only 24% experienced some form of olfactory dysfunction, 76% experienced no olfactory dysfunction, none had anosmia, 18% experienced hyposmia but no parosmia, 3% experienced parosmia but no hyposmia, and 3% experienced both hyposmia and parosmia). As expected, the frequency of combined quantitative and qualitative olfactory dysfunction was significantly lower in the COVID- group compared to the COVID+ group (X² (1, N = 132) = 32.4, p < .0001). This means that COVID-19 increased olfactory dysfunction (parosmia, hyposmia, or parosmia and hyposmia) with 41 percentage points (COVID+: 65%; COVID-: 24%) 15 months after COVID-19 onset.

Discussion

More than a year after recovering from COVID-19, nearly two thirds (65%) of participating healthcare workers still exhibited some form of olfactory dysfunction with more than a third (37%) showing a clinically reduced sense of smell. In comparison, approximately a quarter (24%) of individuals without prior SARS-CoV-2 infection displayed clinically altered sense of smell. Further, amongst the individuals who experienced SARS-CoV-2 infection, about half (51%) experienced parosmia, compared to only 5% in the SARS-CoV-2 naive group; an increase of parosmia due to COVID-19 with 46 percentage points. More specifically, the first wave of COVID-19 nearly doubled the prevalence of olfactory dysfunction in this population from 20% to 36% and increased the prevalence of any form of olfactory dysfunction by a full 41 percentage points compared to individuals whose immune system was never exposed to the virus. Perhaps even more striking, 24% of all tested COVID-19 survivors still experience parosmia 2.6 years after COVID-19 diagnosis, nearly half of which experience medium to severe symptoms.

Olfactory dysfunction is common also in the general population. Before the COVID-19 pandemic, the rate of quantitative olfactory dysfunction was consistently estimated to approximately 20% [16,19]. In our control sample of COVID-19 naïve individuals, we found a similar percentage with 20% of the participants exhibiting quantitative olfactory dysfunction. As outlined above, olfactory dysfunction can have different etiologies, including sinonasal disease, traumatic brain injury, neurodegenerative diseases, and more [20,30]. It is therefore crucial to assess the prevalence of olfactory dysfunction due to COVID-19 against this background of olfactory dysfunction in the population. Our data suggests that COVID-19 roughly doubles the prevalence of quantitative olfactory dysfunction in the general population. A different and bleaker picture emerges when considering also qualitative olfactory dysfunction, i.e., parosmia and phantosmia. These conditions are relatively rare in the general population, which is also reflected by the prevalence of 5% in the COVID-19 naïve group aligning with prior assessments of Swedish samples [31,32]. However, nearly half of the COVID+ group exhibited qualitative olfactory dysfunction on average 15 months after contracting the disease.

Our study shows that 37% of COVID+ individuals still experienced reduction in their olfactory performance, even on average 15 months after COVID-19 onset. This percentage is higher compared to similar studies, see among others [13]. Notably, while there was a significant difference in overall olfactory functions between the two groups, as operationalized by the TDI scores, this difference was restricted to performance on the odor quality discrimination and odor identification subtests, but not in odor detection threshold This finding suggests that a potential underlying driver of the differences is the large difference in parosmia that might distort the odor quality of the odors included in the two-odor subtest.

It is not yet completely understood how a SARS-CoV-2 infection leads to olfactory dysfunction. The leading explanation of the acute olfactory dysfunction seen in patients is linked to the ability of the virus to infect human cells that co-express ACE2 and TMPRSS2 proteins, such as the sustentacular cells of the olfactory mucosa [33]. Upon infection, these cells degenerate, which disturbs the local environment and crucially results in cell death of olfactory receptor neurons and consequently olfactory dysfunction [33,34]. The olfactory system demonstrates however a very good ability to regenerate [35], which might explain why most SARS-CoV-2 infected individuals regain olfactory abilities within weeks following the acute phase [5,9]. It is not yet clear why some individuals do not completely regain their olfactory abilities.

Recent data from the verbal track-and-trace program in the United Kingdom suggests that fewer individuals report subjective olfactory dysfunction after infection with the later Omicron (BA.1. and BA.2) variants than the original virus variants [36]. We do not know what specific virus variants individuals in our sample was infected with or exactly what proportions of virus variants dominated in our specific sample, but random sampling of the Stockholm population at the time indicated that the Wildtype and, to a lesser extent the B.1.1/B.1.1.29 and B.1.1.1/C.14 strains, dominated [29]. Although tentative data suggest that fewer individuals report subjective olfactory dysfunction after Omicron variant infection, these are based on subjective data collected only a day or two after testing positive. Whether potential lower numbers of olfactory dysfunction after Omicron or other later variants could be due to a delay in onset of olfactory dysfunction that in turn might affect long-term outcomes, remains to be determined. Future studies should address differences between virus variants in effects on olfactory function.

A significant strength of the present study is that all participants were continuously serologically monitored from the onset of the pandemic, meaning that it can be firmly established not only that all participants in the COVID+ group had undergone a SARS-CoV-2 infection, but also when. Critically, we can firmly claim that no participants in the COVID- group, a control group from the same cohort, had seroconverted at any point before sensory testing, thereby giving us a true baseline for existing olfactory dysfunction prevalence. Likewise, most studies assessing olfactory dysfunction in a general population suffer from collider bias, i.e., that individuals that experience olfactory problems are more likely to volunteer for the study in the first place, thereby erroneously increasing the prevalence of dysfunction. By using a control group undergoing the same recruitment strategy as the target group and the population being healthcare professionals, possibly better informed and willing to participate in research without personal gain, it can be assumed that collider bias acts in equal strength across groups. It can be argued that healthcare workers, in general, are a group consisting of generally more healthy individuals than the general population. Olfactory problems and their underlying mechanisms therefore could be different than other populations. However, the fact that we obtained identical prevalence of olfactory dysfunction in the COVID- group as the only two previously published prevalence studies in a Swedish sample do not support this notion [31,32]. That said, the strengths mentioned come from the fact that we sampled from a smaller group of closely monitored individuals with a profession that was unfortunately highly taxed for time during a pandemic. This meant that obtaining data from a large and diverse sample was not possible. In addition to the relatively small sample sizes, the COVID+ group had nearly double the participants compared to the COVID- groups, which we aim to statistically control for by using methods such as the Wilcoxon rank sum test. These tests do not assume equal sample sizes and are therefore robust against differences in sample size. Another weakness is that the sex ratio in both groups were skewed with more women participating, which, however, also reflects the underlying population of healthcare workers. Women are also known to be slightly more prone to experiencing long-term effects on olfactory function after upper respiratory infections [37]. Whether the skewed sex ratio of our sample affects the generalization of our results to other populations of healthcare workers is not known.

Conclusion

We show that COVID-19 nearly doubles the already large prevalence of quantitative olfactory dysfunction to approximately 37% in a sample of healthcare workers. Furthermore, about half of the COVID-19 survivors exhibit qualitative olfactory dysfunction. Finally, nearly two thirds (65%) of COVID-19 survivors in this group of healthcare workers exhibit olfactory dysfunction of some form 15 months and 24% reported parosmia 2.6 years after infection. Given the length of time, it is possible that these olfactory problems may not be fully reversible in a plurality of individuals.

Supporting information