Coronavirus disease 2019 (COVID-19) can present with a myriad of symptoms.
1
Guidelines from China, the United Kingdom and Italy had focused screening efforts
on patients with fever and cough, excluding anosmia from similar scrutiny.
1
However, screening of individuals with reported anosmia and dysgeusia has been associated
with a greater likelihood of a positive COVID-19 result than other indicator of an
upper respiratory infection.2, 3, 4 The relative predictive value of presenting COVID-19
symptoms is under current investigation.3, 4, 5, 6 This study seeks to ascertain the
role of smell loss in risk stratification and predicting COVID-19 patients’ prognosis.
Adult COVID-19 positive patients evaluated at a university medical center between
February 1 and April 3, 2020 were identified by an electronic medical records query
and included in our initial series. Complete data on demographic variables, clinical
characteristics, COVID-19 symptoms, COVID-19 treatments, and clinical evaluations
was retrieved. Through a pre-designed screening questionnaire for COVID-19, patients
evaluated by telemedicine, in-person, or at the emergency room were asked about their
symptoms during the history taking, including whether they had acute smell loss. Patients
with incomplete clinical data, or those for whom smell loss was not recorded, were
excluded.
The retrieved information included demographics, BMI, comorbid conditions (asthma,
allergic rhinitis, chronic rhinosinusitis, eczema, food allergy), pre-existing smell
dysfunction, COVID-19 related inflammatory laboratory values (complete blood counts,
c-reactive protein, albumin, creatinine, ferritin and erythrocyte subdimension rate),
COVID-19 outcomes (need for hospitalization, ICU admission, intubation) and development
of acute respiratory disease syndrome. To identify and confirm comorbidities and other
clinical variables, all charts were reviewed by two independent trained researchers
and 20% of the charts were randomly checked by the principal investigator. Data points
with lack of concordant information were reviewed again by an independent investigator,
and if needed excluded from analysis.
SPSS v23 (SPSS, Inc., Chicago, IL, USA) was used for all analyses. Continuous variables’
results are presented in the text as mean ± SD, unless otherwise specified, and were
compared using parametric if normally distributed (Student’s t test). The χ2 test
was performed to analyze the correlation between categorical parameters. Logistic
regression was conducted to calculate the odds ratio (OR) of smell loss in association
with nominal dependent variables such as pre-existing comorbidities as well as COVID-19
outcome adjusted for possible confounders (demographics and BMI). The adjusted ORs
are presented with their 95% confidence intervals (CIs). Analysis of covariance (ANCOVA)
was conducted to compare the adjusted means of continuous variables such as laboratory
values in association with smell loss, adjusting for demographics and BMI. This research
study was approved by the Institutional Review Board.
The initial series consisted of 1013 patients who were evaluated and tested positive
for COVID-19. Sufficient data on smell loss, demographic variables, comorbidities,
and outcomes was available in 949 patients (93.7%) who were included for analysis.
The cohort consisted of 55.2% female patients, with a mean age ± standard deviation
(SD) of 48.42±15.67 years. In this series, 54.3% of patients were African-American
or Black, 25.0% were non-Latino White, 22.9% were Latino, and 14.3% were identified
as other race/ethnicity.
Overall, 198 (20.9%) patients reported smell loss during their initial evaluation
for COVID-19. Smell loss was significantly associated with younger age, female gender,
and higher BMI. The mean age was 46 versus 49 years in those with and without smell
loss, respectively (p=0.02); 64.7% of subjects with smell loss vs 52.8% of those without
smell loss were females (p= 0.003); and mean BMI was 33.6 vs 31.5 in those with and
without anosmia, respectively (p=0.001). There was a significant association between
smell loss and history of pre-existing smell dysfunction (OR, 4.66; 95% CI, 2.07-10.46),
allergic rhinitis (OR, 1.79; 95% CI, 1.12-2.87), and chronic rhinosinusitis (CRS)
(OR, 3.70; 95% CI, 1.29-10.67) compared to patients without smell loss.
Sufficient data on laboratory markers was available for 419 (41.8%) patients. Compared
to patients without smell loss, patients with smell loss demonstrated less lymphopenia
(the mean ± SD of lymphocyte count was 1.84 ± 3.69 vs 1.11 ± 0.81 in those with and
without smell loss, p = 0.001) and higher albumin counts (3.02 ± 0.83 vs 2.77 ± 0.83,
p = 0.02). Other laboratory values and inflammatory markers were not associated with
smell loss among COVID-19 positive patients. These results did not change after adjusting
for demographics and BMI.
Smell loss was also significantly associated with decreased hospitalization (OR, 0.69;
95% CI, 0.47-0.99), ICU admission (OR, 0.38; 95% CI 0.20-0.70), intubation (OR, 0.43;
95% CI, 0.21-0.89), and ARDS (OR, 0.45; 95% CI, 0.23-0.89) after adjustment for demographics
and BMI (see Table 1
). These results remained significant after further adjustment for allergic rhinitis
and CRS.
Table 1
Preexisting conditions and COVID-19 related outcomes in 949 COVID-19 patients in association
with smell loss
Conditions
Number of cases with condition among the series
Odds ratio (95% confidence interval) of having smell loss in patients with condition
compared to those without the condition
Adjusted p.value ¥
History of past smell dysfunction
27
4.66 (2.07-10.46)
<0.001∗∗∗
Allergic rhinitis
101
1.79 (1.12-2.87)
0.02∗
Food allergy
71
1.64 (0.95- 2.83)
0.08
Atopic dermatitis (eczema)
42
1.22 (0.59-2.50)
0.60
Asthma
243
1.18 (0.82-1.70)
0.36
Chronic rhinosinusitis
15
3.70 (1.29-10.67)
0.02∗
GERD
60
0.67 (0.29-1.58)
0.36
Diabetes
243
0.86 (0.57-1.29)
0.46
Hypertension
391
1.14 (0.77-1.68)
0.52
Emergency room visit for COVID-19
520
0.83 (0.59-1.16)
0.28
Hospitalized
311
0.69 (0.47-0.99)
0.04*
ICU admitted
131
0.38 (0.20-0.70)
0.002*
Intubated
86
0.43 (0.21-0.89)
0.02*
ARDS
93
0.45 (0.23-0.89)
0.02*
¥ = Odds ratios and adjusted p.values are calculated by logistic regression adjusting
for age, gender and BMI.
*: p.value<0.05
***: p.value<0.001
Our data implicates smell loss as an independent positive prognostic factor of a less
severe COVID-19 infection. It was significantly associated with decreased hospitalization,
ICU admission, intubation, and ARDS rates compared to the absence of smell loss. In
further support, a smaller studies of 169 and 34 COVID-19 positive patients showed
an association between anosmia with outpatient care as opposed to hospitalization
7
. Our data aligns with these findings. Additionally, smell loss was associated with
less lymphopenia and higher levels of albumin, suggesting a less severe reaction to
COVID-19 in patients with smell loss compared to those with intact smell.
A history of pre-existing smell dysfunction, allergic rhinitis, or chronic rhinosinusitis
(CRS) was associated with greater chance of acute smell loss in patients with COVID-19.
However since most patients who experience smell loss in the setting of COVID-19 report
return of smell with clinical resolution of illness
8
and an initial neuroimaging study appears to show an absence of acute visible size
changes to the neural olfactory system
9
, COVID-19 is not associated with permanent anosmia. Positive and negative predictive
values could not be calculated because the basal rates of hyposmia and anosmia and
the prevalence of COVID infection and each infected subject individual phase of illness
are not established for the studied population, but have been examined in further
detail elsewhere
6
. Female gender, lower age, higher BMI, history of previous smell loss, and pre-existing
allergic rhinitis and chronic rhinosinusitis appeared as important predictors of smell
loss in the setting of COVID-19 infection. The main limitations of our study were
its retrospective nature, subjective nature of smell loss, and focused nature of the
data collection, which did not include subjects’ current medications.