Symptoms of depression change with olfactory function

Olfactory loss is associated with symptoms of depression. The present study, conducted on a large cohort of mostly dysosmic patients, aimed to investigate whether improvement in olfactory performance would correspond with a decrease in depression severity. In 171 participants (157 dysosmic), we assessed olfactory function and severity of depression before and after an average interval of 11 months, with many patients showing improvement in olfactory function. Separate analyses were conducted for (a) the whole group of patients and (b) the group of dysosmic patients using both classic and Bayesian approaches. For odor identification, Student t test demonstrated that the whole sample improved consistently, especially within the group of dysosmic patients. The dysosmic group also improved in odor threshold and overall olfactory function. Pearson correlation showed that an increase in olfactory function was associated with a decrease in depression severity, particularly in dysosmic patients. To conclude, the present results indicate that symptoms of depression change with olfactory function in general and odor identification in particular.

www.nature.com/scientificreports/ All participants visited the Smell and Taste Clinic of the Department of Otorhinolaryngology at the TU Dresden because of an olfactory disorder. The duration of treatment varied from 1 up to 16 months (10.9 ± 3.2) and consisted of different medical strategies presented in Table 1. Different etiologies of olfactory loss were present ( Table 1).
The retrospective study was performed according to the principles of the Declaration of Helsinki on biomedical research involving human subjects. It was approved by the Ethics Committee at the Medical Faculty of the TU Dresden (EK number 251112006) covering anonymized retrospective and pooled analyses. All participants consented to the examinations.
Olfactory testing. Detailed results of olfactory testing are presented in Table 2. The testing was conducted twice, during the first and the second visit to the Clinic. 39-40 patients did not attend visit 2-they were followed up instead with a telephone consultation.
Olfactory testing was performed by means of "Sniffin' Sticks" 49,50 and consisted of tests for odor threshold (rose-like odor, phenylethylalcohol; PEA), odor discrimination, and odor identification. Results of the 3 subtests were presented both separately for threshold (T), with a range between 1 and 16, discrimination (D), with a range between 0 and 16 and identification (I) score, with a range between 0 and 16, and finally as a sum of the results (TDI), with the final score ranging between 1 and 48 points 50 . If the TDI score was 31 or higher, the patient was regarded as normosmic (n = 14); with a score below 31, the patient was regarded as dysosmic (n = 157) 50 .

Neuropsychological testing.
Detailed results of the depression scale are presented in Fig. 1. Like olfactory testing, depression testing was conducted twice, during the first (M = 14.32, SD = 9.89) and the second (M = 14.36, SD = 10.45) visit to the Clinic. In case of patients who were not able to come for the second visit, the measure was conducted via phone.
The General Depression Scale-long form 51,52 ("Allgemeine Depressionsskala"-Langform; ADSL) was employed to gauge the severity of depression. The scale is a self-assessment tool that consists of 20 questions about emotional, motivational, cognitive, somatic and motor/interactional complaints, such as, e.g. 'During the last week I hardly had an appetite'; 'During the last week I felt as good as others'; 'During the last week I was scared' . Due to its time-saving and cost-effective applicability, the ADSL is a practical method 51 . Patients are  Statistics. Separate analyses were conducted for (a) the whole group of patients, with the specification of normosmic patients and (b) the group of dysosmic patients. Possible differences between the first and the second measure in (a) olfaction and (b) depression were investigated by means of paired samples two-way t test. Additionally, Repeated Measures Anova was conducted to assess possible differences between dysosmic and normosmic patients, with session (before-after) as within factor and olfactory groups (normosmic vs. dysosmic) as between-subject factor. Independent samples t test was used to examine possible differences in depression scores between the normosmic and the dysosmic patients.
Pearson correlation was employed to investigate the relationship between the improvement in olfactory function and depression severity. Improvement in odor function was measured so that the first olfactory measure, separate for T, D, I and TDI was subtracted from the second olfactory measure. Improvement in depression was calculated in reverse: the second depression score was subtracted from the first one. This was done so that positive differences indicate improvement in olfactory function ion and depressive symptoms, respectively. Additionally, to confirm the obtained results, Principal Component Analyses was conducted (see Supplementary materials).
All the analyses were conducted twice, firstly with classical p-values and secondly with Bayesian statistics 53 . The Bayes Factor (B) is a method that weighs evidence and shows which of two hypotheses is better supported and to what extent. Adopting the B in statistical inference, it can be examined whether the data provided robust support for the null hypothesis, the alternative hypothesis, or whether the analysis is inconclusive and more data need to be collected to provide clear evidence 54 . Furthermore, Bayesian statistics are resistant to multiple comparisons.
Data are presented as mean values (± standard deviation). Statistical analyses were performed using JASP v. 0.11.1 (Univ. of Amsterdam, The Netherlands: www. jasp-stats. org), with p < 0.05 set as the level of significance. The effect sizes are accompanied by their 95% confidence intervals.   Furthermore, no difference on the assumed significance level was found between the dysosmic and normosmic group in terms of depression severity, either for the first measure (but see size effect: t = 1.49, p = 0.14, Cohen's d = 0.42) or the second one (t = 0.4, p = 0.69, Cohen's d = 0.11). The Bayesian factor indicated that the severity of depression in the second measure was the same in both groups while in the first measure more data would be needed to support or reject this hypothesis (B01 for the first measure = 1.44; for the second measure = 3.35).  www.nature.com/scientificreports/ was, respectively, strongly and moderately confirmed by the Bayesian factor (for "TDI score" B10 = 12.8, for I score: 5.51 (Supplemental Table 1).

Dysosmic patients.
For the group of dysosmic patients, no correlation was observed between the change in depression severity and the change in odor discrimination (r = 0.09, p = 0.35), or odor thresholds (r = 0.05, p = 0.60). Instead, changes in depression severity correlated positively with the change in general olfactory function (r = 0.28, p < 0.001) and in odor identification (r = 0.25, p = 0.003) (Fig. 4). This was confirmed by Bayesian statistics, strongly for general olfactory function (B10 = 28.5) and moderately for odor identification (B10 = 8.8) (Supplemental Table 2).

Discussion
In order to explore the idea that depressive symptoms can benefit from improved olfaction via repeated exposition to odors, we examined whether olfactory improvement is associated with a reduction in depression severity. Specifically, in a large sample of primarily dysosmic patients, we sought evidence for an association between change in depression severity (measured by ADSL scale 52 ) and variation in olfactory function (namely odor threshold (T), discrimination (D), identification (I), and general olfactory function (TDI) as measured by the Sniffin Sticks test battery 49,50 ). In accordance with our assumption, the present results show that an increase in olfactory function corresponded moderately to strongly with the decrease in depressive symptoms, which was particularly notable in the group of dysosmic patients. Furthermore, the whole sample improved consistently, especially within the group of dysosmic patients in terms of I, and the dysosmic group also improved in terms of T and TDI. The majority of the participants in the present study were dysosmic due to viral infections, sinonasal disease, and traumatic brain injury, which are the most typical causes of olfactory dysfunction 13,55-57 . Sinonasal disease may lead to a gradual loss of olfactory function generated by nasal obstruction blocking access of odors to the olfactory cleft and chronic inflammation of the olfactory epithelium 58 . In addition, infections have been demonstrated to result in damage to the olfactory mucosa, thereby disabling the function and regeneration of olfactory receptor neurons 59,60 . Finally, head trauma may disrupt olfactory pathways or damage olfactory brain areas 61,62 .
Previous studies demonstrated that dysosmic patients can benefit from OT to improve their sense of smell and reduce their symptoms of depression 30,31 . Odor identification is an olfactory dimension with a solid cognitive association 63 that has been consistently reported to be impaired in depression [17][18][19][20][21] , probably because of the cognitive impairment in recurrent depression 19 . Specifically, reduced attention to and memory of olfactory stimuli www.nature.com/scientificreports/ may lead to decrease in odor identification 2 . Furthermore, odor identification is also more subject to complex processing compared to odor thresholds 63 and thus may show more overlap with areas that are involved in the processing of emotions, and hence, depression. Bestgen and colleagues 64 showed that odor identification was predicted by odor emotional quality, and a number of studies indicated its relation with depression [17][18][19][20][21] .
The present finding indicating a relationship between odor identification and depression severity in dysosmic patients may also be linked to the large improvement in odor identification in this group, which was five times greater than in the whole sample. We presume that better general olfactory and odor identification performance may lead to cognitive and affective improvement.
Notably, the majority of the sample in the present study was not clinically depressed, as indicated by the results of the ADSL scale 52 . However, according to the previously mentioned scale, a significant percentage of the present sample (one-third of the entire sample) was classified as depressed. That could probably explain why depression severity before and after medical olfactory treatment described in Table 1 remained the same, contrary to previous studies involving OT 30,31 . In this context, strong and moderate relationships obtained between change in TDI, I, and depression severity suggest that changes in olfaction can be related even to very subtle changes in mood.
The relationship between changes in olfaction and mood is also underlined by the difference between correlation results in both groups. It is worth noting that in the group of dysosmic patients, both improvement of general olfactory function (TDI) and odor identification score (I) corresponded with a decrease in depression severity to a higher degree than in the whole sample of patients which included several normosmic participants.
Olfactory impairment has been demonstrated to coexist with depression both in the general population 1,6,10 and in specific groups, such as older adults [ 15,16 , 65,66 ; but see also: 67 ]. This age group is particularly prone to reduced olfaction 68 . Hence, this group appears to be specifically exposed to the consequences of olfactory dysfunction. In a very recent study, Qazi and colleagues 66 demonstrated that olfactory impairment predicted the occurrence of depressive symptoms in a group of older adults. In line with this evidence, the present study results, where almost half of the participants were older than 65 years old, show that controlling for olfactory function might be an efficient tool to mediate mood changes.
The present study is not free from limitations. Firstly, because of the outbreak of the pandemic COVID-19, 23% of patients did not participate in person in the follow-up session and were examined only by a phone interview. Secondly, it should be noted that correlation results between the change in odor identification and depression severity were moderate according to Bayesian classification. Nevertheless, depression severity is a complex issue with a number of varying factors related to it 69 , and odor identification is only one among them. Future studies in this field should further examine this issue. Also, patients who suffer emotionally from their smell impairment are more likely to visit specialized centres on smell and taste and to return for re-testing. Thus, the present study sample probably showed more depressive symptoms compared to other people with smell impairment who do not search for clinical counseling 70 .
To conclude, the present study results conducted on a large cohort of mostly dysosmic patients indicate that olfactory function in general (TDI), and odor identification (I) in particular, may be used as an indicator of subtle mood changes. Further, the results suggest an improvement of symptoms of depression with an improvement in the sense of smell.

Data availability
The datasets analyzed during the current study are not publicly available due to the privacy of the participants but are available from the corresponding author on reasonable request.