Olfaction and thyroid hormones in patients with subjective cognitive decline, non-amnestic and amnestic mild cognitive impairment

Data were gathered from patients who consulted the Memory Outpatient Clinic regarding memory concerns. The patients were assigned either by a physician or through self-referrals.

The participants underwent physical, neurological and neuropsychological examinations, standard laboratory testing (including serum thyroid parameters), and neuroimaging to rule out cerebrovascular diseases. Neuropsychological tests included the mini-mental state examination (MMSE), Vienna visuo-constructional test (VVT.03), vocabulary test (Wortschatz Test, WST-IQ), Beck Depression Inventory (BDI-II) and the Neuropsychological Test Battery Vienna (NTBV). Sociodemographic characteristics, such as age, sex, and education in formal school years, were noted. The OF was measured using Sniffin’ Sticks© of Burghart Instruments and the Assessment of Self-reported Olfactory Functioning (ASOF) test questionnaire.

Patients were categorized into SCD, naMCI, and aMCI based on clinical examination, neuropsychological test scoring, and classification guidelines using criteria presented by Jessen et al. [16] for SCD and Petersen’s criteria [17] for the diagnosis of MCI patients (aMCI and naMCI).

Based on serum TSH levels, patients were further categorized into hyperthyroidism, euthyroidism, and hypothyroidism subgroups. Participants were included if the age at the time of cognitive and olfactory testing was ≥50 years, and their blood samples for TSH were available within ±3 months of the neuropsychological assessment.

Participants were excluded if they fell into one or more of the following categories: age < 50 years, missing TSH values, presence of overt neurological disease, other types of dementia and cognitive impairment, evidence of stroke as determined by neuroradiological and clinical examinations, history of traumatic head injury, psychiatric problems causing pseudodementia, and medical conditions associated with severe cognitive impairment.

Out of the initial 741 protocols 246 (33.2%) were excluded considering the inclusion and exclusion criteria, particularly TSH parameter completeness. This resulted in a study collective of N = 495 patients (81 SCD, 231 naMCI, and 183 aMCI), as shown in Fig. 1.

Thyroid parameters were extracted from laboratory tests performed on patients after consultation at the outpatient clinic. Serum TSH, total thyroxine (T4), free thyroxine (fT4), total triiodothyronine (T3), and free triiodothyronine (fT3) levels were determined based on standard laboratory values and reference ranges of the General Hospital of Vienna (AKH Wien), where the electrochemiluminescence immunoassay method (ECLIA) was used. The reference ranges are 0.27–4.2 µIU mL⁻¹ (TSH), 0.8–1.8 ng mL⁻¹ (T3), 58–124 ng mL⁻¹ (T4), 2.15–4.12 pg mL⁻¹ (fT3), and 0.76–1.66 ng dL⁻¹ (fT4).

As TSH is the most sensitive and commonly used biomarker in determining thyroid function and as not all participants had fT3 and fT4 values available, thyroid function was assessed using only TSH levels (TSH ⇧ represents hypothyroidism, TSH ⇩ is hyperthyroidism, and TSH ⇨ is euthyroidism).

The OF was assessed using Sniffin’ Sticks© and Assessment of Self-reported Olfactory Functioning (ASOF) test for objective and subjective measurements, respectively. The Sniffin’ Sticks© test by Hummel et al. [18] is an objective nasal chemosensory performance test consisting of three subtests. This study only used the odor identification subtest and 16 common odors were presented for 3s via a felt-tip pen. The subjects were asked to identify the correct odorant from a list of four descriptors in a single choice format [18]. Scores range from 0 to 16, with ≥11 indicating intact OF and <11 reflecting hyposmia in patients over 50 years of age [19].

The ASOF test developed by Pusswald et al. [20] is a 12-item self-administered questionnaire assessing the subjective OF and its impact on quality of life. It comprises three subtests: the Subjective Olfactory Capability Scale (SOC) rating the overall olfactory ability on a scale of 0–10, the Smell-Related Problems Scale (SRP), including 5 frequency questions and the Olfactory-related Quality of Life Scale (ORQ) consisting of 6 impairment quality questions, both using a common 5‑point Likert rating scale. Cut-off scores, based on healthy control mean scores, identify decreased olfactory abilities in SOC scores ≤3, SRP scores ≤2.9, and ORQ scores ≤3.7. [20] The ASOF can be obtained from www.​psimistri.​com.

To evaluate the cognitive state of the participants, the following neurocognitive tests were used: the MMSE [21], VVT (version 3 0) [22] to assess visuo-constructive function, WST-IQ [23] to assess verbal intelligence and language comprehension and the NTBV, [24] a standardized instrument for the early detection of objective cognitive impairment and dementia consisting of 24 subtests, which represent the following cognitive domains: attention, executive function, language, psychomotor speed and memory. To assess the severity of depressive symptoms, the revised BDI-II [25] was used. The VVT3.0 and the NTBV can be obtained from www.​psimistri.​com.

Descriptive and inferential analyses were performed using the statistical software IBM SPSS® 29.0.0 and Excel® for MacOSX. The significance level was set at alpha = 5% according to the type I error.

Metric parameters, at least interval-scaled ones, were characterized via mean (M) ± standard deviation (SD), minimum (min) and maximum (max). For skewed data distribution, the median (Mdn) and interquartile range (IQR, percentage range of 25–75%) were assessed. For the description of nominally scaled parameters, frequencies (n) and percentages (%) were presented. If appropriate, 95% confidence intervals (CI) were calculated to specify a range estimate.

In the inferential section, analyses of variances (ANOVA) were performed to test for differences in metric (at least interval-scaled) parameters between the three cognitive impairment groups. Additionally, the homogeneity of variances had to be considered using Levene’s test. In the case of heterogeneity, Welch-ANOVA was conducted. If the data were skewed, the nonparametric Kruskal-Wallis procedure was used and post hoc pairwise comparisons were conducted using the Mann-Whitney U test.

Regarding NTBV subtests, a principal component factor analysis (PCA) was performed to achieve a dimensional reduction of 24 subtests into a smaller number of independent components. This procedure was conducted utilizing an orthogonal Varimax rotation. This approach reduced redundancies and enabled weighted independent z-factor scores (μ = 0, σ = 1). It also allowed an adapted and standardized integration of NTBV subtests. The adequate interpretation of this dimensional reduction required two key values: the communality (h_i² ≤ 1, sum of squared factor loadings regarding one subtest considering the extracted components) and the eigenvalue (λ ≥ 1, sum of squared factor loadings regarding one extracted component considering all subtests).

As baseline characteristics, the data on demographics, thyroid, olfactory, and neuropsychological parameters are presented in Table 1. The proportion of female participants was 52.3% (95% CI 47.9–56.7%). The mean age of participants at testing was 69.2 ± 9.3 (min: 50.1; max: 90.4) years and no significant differences in age between cognitive subgroups (p = 0.195) and sex (p = 0.673) could be assumed. Similarly, no significant difference between cognitive subgroups regarding years of schooling was observed (p = 0.731).

Regarding thyroid parameters, the comparison of metric laboratory values for all five hormones TSH (µIU mL^-1), T3 (ng mL^-1), T4 (ng mL^-1), fT3 (pg mL^-1), fT4 (ng dL^-1) showed insignificant differences between the three cognitive groups, p ≥ 0.199. Being the most relevant parameter, TSH value distribution in the three cognitive subgroups is displayed in Fig. 2.

Furthermore, the classification of hypothyroidism, euthyroidism, and hyperthyroidism was determined based on the TSH reference ranges. The distribution of functions in the three cognitive groups is provided in Fig. 1. Accordingly, the proportion of euthyroidism was 98.8% for SCD, 93.9% for naMCI, and 90.2% for aMCI. Overall, 93.3% of all participants had normal thyroid functions.

Regarding olfaction, substantially significant differences in objective olfactory performance (Sniffin’ Sticks) were found between the three cognitive groups; p < 0.001; post hoc pairwise comparisons revealed that the aMCI group showed weaker olfactory performance than SCD and naMCI, p < 0.001, considering Bonferroni adjustment, indicating small effects, r = 0.23 [25]; however, concerning subjective olfactory assessment via the ASOF test (SOC, SRP, SRQ), pairwise post hoc comparisons revealed no substantial differences between cognitive subgroups. Accordingly, the extent of agreement according to rank correlation between objective (Sniffin’ Sticks as criterion) and subjective odor performances for SCD subjects was between r_s of 0.40 and 0.45 for naMCI and between r_s of 0.18 and 0.36 for aMCI.

Additionally, the proportion of hyposmic participants, taking an odor identification cut-off <11 into account, was 22.2% (13.2%; 31.3%) for SCD, 26.0% (20.3%; 31.6%) for naMCI, and 48.1% (40.8%; 55.3%) for aMCI, indicating a decreased odor performance regarding aMCI patients. In contrast, the ASOF subscales SOC, SRP, and ORQ showed no significant distributional differences comparing cognitive groups. Additionally, the proportions of subjective hyposmia ranged between 4.9% and 12.0%.

The consideration of OF (Sniffin’ Sticks <11 hyposmic vs. normal) and thyroid function did not reveal a significant increase in hyposmia with respect to hyperthyroidism and hypothyroidism, implying that these two entities are considered independent of each other. Similarly, the correlative relationships of Sniffin’ Sticks and ASOF subtests (SOC, SRP, ORQ) with TSH values overall (N = 495) and in the three diagnostic subgroups revealed only weak coincidences (r_{s of} −0.19 to +0.12).

Regarding the relation between the other thyroid hormones (T3, T4, fT3, and fT 4) and Sniffin’ Sticks as well as ASOF subscales, in the overall study collective (N = 495), only a weak negative significant relation (p = 0.039) was observed for fT4 (ng dL⁻¹) and Sniffin’ Sticks (r_s (n = 127) −0.18), implying that higher olfactory performance was associated with lower fT4 levels.

Neurocognitive performance was assessed via MMSE, VVT 3.0, WST-IQ, and NTBV and depression via BDI-II, as shown in Table 1. The NTBV-24 subtest results were subject to a principal component analysis (PCA) using the varimax rotation method according to Kaiser [37], achieving a dimensional reduction. The information criterion, with Kaiser–Meyer–Olkin test (KMO) = 0.838, indicates a substantial extent of information for carrying out an explorative factor analysis. These extracted components cover a total of 71.9% explained variance. The following meta-key terms can be proposed to label the six dimensions (inspired by the original NTBV domain names): (F1) perceptual functions and attention, (F2) concentration, (F3) verbal memory, (F4) executive functioning skills, (F5) verbal fluency, and (F6) nonverbal fluency (Table 2). Weighted and independent z-standardized factor scores were derived from these six dimensions, which are available for further analyses.

Finally, a hierarchical modelling for influencing factors was conducted using an adjusted multiple linear regression to predict olfactory performance assessed via Sniffin’ Sticks identification subtest.

The development of coefficients by adjusted modelling for Sniffin’ Sticks olfactory identification is presented in Table 3. The tolerance value of all predictors was ≥0.61, indicating no striking multicollinearity. A normal distribution of standardized z-residuals and homoscedasticity could be assumed. No noticeable autocorrelations of the residuals were observed, as indicated by Durbin-Watson’s coefficient = 1.91.

The global model summary revealed an explanatory value of R²_adj. = 22.5%, p < 0.001, indicating a significant contribution of predictors for olfactory identification (Sniffin’ Sticks). Increasing age, as seen in the last fifth model block, with a small weight (β = −0.204, p < 0.001) appears as a risk factor for decreased olfactory identification, whereas a better performance of the NTBV dimensions verbal memory (F3) with a moderate weight (β = +0.327, p < 0.001), concentration (F2) with a small weight (β = +0.119, p = 0.004) provide increased smell identification. All other independent variables and covariates were not significantly contributing to predicting objective olfactory performance. In particular, TSH did not reveal a substantial weight.