Development of a Korean Culture-Friendly Olfactory Function Test and Optimization of a Diagnostic Cutoff Value

Article information

Clin Exp Otorhinolaryngol. 2020;13(3):274-284

Publication date (electronic) : 2020 July 17

doi : https://doi.org/10.21053/ceo.2020.00864

Jong-Gyun Ha¹, Jinwon Kim¹, Jae Sung Nam¹, Jeong Jin Park¹, Hyung-Ju Cho^1,2,3, Joo-Heon Yoon^1,2,3, Chang-Hoon Kim^,1,2,3,4

¹Department of Otorhinolaryngology, Yonsei University College of Medicine, Seoul, Korea

²The Airway Mucus Institute, Yonsei University College of Medicine, Seoul, Korea

³Korea Mouse Sensory Phenotyping Center, Yonsei University College of Medicine, Seoul, Korea

⁴Taste Research Center, Yonsei University College of Medicine, Seoul, Korea

Corresponding author: Chang-Hoon Kim Department of Otorhinolaryngology, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun-gu, Seoul 03722, Korea Tel: +82-2-2228-3609, Fax: +82-2-393-0580 E-mail: entman@yuhs.ac

Received 2020 May 13; Revised 2020 June 3; Accepted 2020 June 17.

Abstract

INTRODUCTION

Olfaction determines the ability to detect various odors, including those associated with hazards such as spoiled food, gas leaks, and fires. Olfaction is related to quality of life [1] and survival [2]. It is also a prodromal marker for the early diagnosis of neurodegenerative and psychiatric disorders such as Parkinson disease, Alzheimer disease, and major depression [3-5]. Thus, there is an urgent need to accurately assess an individual’s sense of smell appropriately. Many psychophysical olfactory tests have been developed and used in research and/or clinical settings; these evaluate one or more olfactory components of the detection threshold, discrimination, and identification. The detection threshold, discrimination, and odor identification are affected by low, middle, and high levels of olfactory processing, respectively, and require different levels of cognition [6]. In particular, odor identification may require prior exposure to culturally-specific odor stimuli [7]. Therefore, many culture-friendly odor identification tests have been specifically developed or modified and validated [8].

The YSK olfactory function test (YOF test; Kimex Co., Suwon, Korea) is a new olfactory function test that uses culturally familiar odorants to Koreans. The 12 odorants of the identification test were selected to include all eight major chemical functional groups.

MATERIALS AND METHODS

YOF test protocol

The YOF test was designed considering the chemical and structural characteristics of the odorants and cultural experience and safety of participants. The YOF test included three subtests, namely, threshold, discrimination, and identification tests. Score range of the YOF subtests were 1–12 for threshold test, 0–12 for discrimination test and identification test, and 1–36 for total TDI score. The commercially available YOF test provided all odorants as felt-tip pens (Fig. 1). The test procedure required covering the subject’s eyes, placing the pens 2 cm away from the nostrils, and subsequently smelling each pen for 3 seconds. All the YOF test procedures were performed in the olfactory function test room equipped with an air ventilation hood.

Fig. 1.

The YSK olfactory function test is composed of three subtests of threshold, discrimination, and identification, which each have 12-point scales.

Odor thresholds test (T)

The odorant for the threshold test was 2-phenylethyl alcohol (PEA), a pleasant, rose-like odorant having minimal trigeminal nerve stimulation [9]. PEA is one of the most frequently selected test odorants for odor threshold tests. Our protocol required dilution of PEA in a geometric series consisting of 12 steps, with a dilution ratio of 1:2 beginning with 10% (v/v) and using dipropylene glycol (DPG) as a solvent. The test method followed a three alternative forced-choice and seven-reversal initially ascending single-staircase procedure, as described by Doty et al. [10]. When starting the threshold test, the subject first smells the strongest concentration of PEA pen to recognize the rose-like odor of PEA. And then, a triplet of odorant pens was presented to each subject in a randomized order, with only one pen containing PEA at a particular dilution and the others containing only solvent (DPG). Subjects were directed to choose a rose-scented pen from the triplet. If an incorrect response was given in any trial beginning with the lowest concentration, the staircase was moved upward until the correct pen was detected at a given concentration in two consecutive trial sets. If subjects gave the correct answer twice in a row, the process went down onestep, and they were directed to choose the correct smell again. The test was conducted until 7 turning points occurred. The geometric mean of the last 4 staircase reversal points in a total of 7 was used as the estimated detection threshold value.

Odor discrimination test (D)

The odor discrimination test incorporated the triplet forced test to assess capability for distinguishing different odorants. The 12 pen triplets for each step included one pen containing one target odorant and two others containing a non-target odorant (Table 1). The target and non-target odorants were set to a similar intensity for the test. Since the chemical complexity of the odorants can affect the sense of smell, the odorants of the discrimination test used a single chemical odorant rather than a mixture [11]. Subjects were directed to select one from each pen triplet. Each step was designed to be easily distinguishable by at least two of three average individuals.

Table 1.

Odorants used in the YOF discrimination test and the correct answer rates for the normosmia group

Odor identification test (I)

In the preliminary study to select odorants for identification test, 113 (male, 71; female, 42; mean age, 38.5 years) subjects with normal olfactory function underwent a test with 32 candidate odorants. These odorants were presented as liquid in brown bottles 3 cm away from subjects. Subjects smelled the odorants by cupping their hands above the bottles and wafting the air toward their faces. Fourteen odorants showed results with >90% correct answers in the preliminary test. Finally, considering cultural familiarity and including eight major chemical functional groups, 12 odorants were selected. The YOF identification test comprised eight universal odorants and four Korean culture-friendly odorants. In addition, it covered eight major chemical functional groups for odorants, namely, ketone, terpene, aldehyde, aromatic, alcohol, ester, acid, and amine (Table 2).

Table 2.

Composition of odorants for the YOF identification test

The odor identification test was designed as a multiple forced choice from four choices (Table 3). All odorants of identification test were presented as a felt tip pen. Subjects had to make one choice from four descriptors after smelling the pen for 3 seconds. From step to step, there was 30 seconds of interval to prevent olfactory fatigue.

Table 3.

Odorants used in the YOF identification test, their distractors, and the correct answer rates for the normosmia group

RESULTS

Demographic data and mean olfactory function test scores

The demographic data and mean values for each T/D/I subtest and the TDI score from each group are presented in Table 4. The sex ratios (male-to-female) within each group were significantly different (normosmia group, 335:207; hyposmia group, 245:227; anosmia group, 54:59; P<0.001 with the Pearson chi-square test). The rate of anosmia was higher in females, but this difference was not significant (P=0.069). Mean age was significantly higher in the anosmia group than in the other groups (normosmia group, 47.0±16.9 years; hyposmia group, 49.6±16.3 years; anosmia group, 52.9±14.1 years; P<0.001, with one-way ANOVA). According to the post hoc Tukey analysis, significant differences were found between the mean age of the normosmia group and that of the hyposmia (P=0.031) and anosmia groups (P<0.001), but not between the mean age of the hyposmia group and the anosmia group (P=0.131). The mean values of each T/D/I subtest and the total TDI scores among the normosmia, hyposmia, and anosmia groups were significantly different in both the YOF test and KVSS-II (P<0.001).

Table 4.

Demographic data and descriptive statics for the YOF test and the KVSS-II by self-assessed olfactory function

Comparison between the YOF test and KVSS-II in the normosmia group

The score distributions (Fig. 2A-D) and olfactory function by age (Fig. 2E-H) of each T/D/I subtest for threshold, discrimination, and identification, as well as the total TDI score, are shown in Fig. 2. The results for threshold (β=–0.025, R²=0.031, P<0.001), discrimination (β=–0.046, R²=0.088, P<0.001), identification (β=–0.046, R²=0.060, P<0.001), and total TDI score (β=–0.115, R²=0.081, P<0.001) consistently showed significant negative correlations with age in the logistic regression analysis. For greater clarity, the distribution of the YOF test and KVSS-II scores (total TDI score and T/D/I subtest scores) in the normosmia group were graphically expressed through the score ratio (0 to 1, test score/score scale) (Supplementary Fig. 1).

Fig. 2.

Score distributions and scatter plot by age of the total TDI score and each T/D/I subtest. The results of the YSK olfactory function (YOF) test of the normosmia group showed grossly normal distributions (A, B, D) except for the identification test (C, intentionally skewed to the right). The horizontal scale of the threshold graph (A) is expressed in units of 1 point for simple visualization. For total TDI score and all results of each T/D/I subtest, olfactory function significantly decreased with increasing age of the subjects (E-H; P<0.001, respectively). T, threshold; D, discrimination; I, identification.

Threshold

The threshold score of the YOF test followed a normal distribution (Fig. 2A). The mean T value of the YOF test was 4.6±2.3 on a 12-point scale, whereas that for the KVSS-II was 6.8±2.6 on a 16-point scale in the normosmia group (Table 4). There was a significant difference in the mean score ratio of the threshold test between the YOF test (mean, 0.38; standard error of the mean [SEM], 0.01) and the KVSS-II (mean, 0.42; SEM, 0.01) (P<0.001) (Supplementary Fig. 1). This statistical difference was caused by differences in the content of each olfactory function test (e.g., test odorant, concentration, score scales, etc.).

Discrimination

The percentage of correct answers for each triplet ranged from 65.6% to 88.1% (Table 1). The mean D score of the YOF test for the normosmia group was 8.6±2.1 on a 12-point scale, whereas that for the KVSS-II was 11.5±2.7 on a 16-point scale (Table 4). There was no significant difference in the mean score ratio of the discrimination test between the YOF test (mean, 0.71; SEM, 0.01) and the KVSS-II (mean, 0.72; SEM, 0.01) (t=0.23, P=0.820) (Supplementary Fig. 1).

Identification

The percentage of correct identifications for the universal odorants ranged from 80.7% to 97.2% (mean, 92.3%; SD, 5.4%) and that for the Korean culture-friendly odorants ranged from 94.0% to 98.6% (mean, 96.2%; SD, 2.1%) (Table 3). However, there was no significant difference in the mean percentage of correct identifications between the universal odorants and Korean culture-friendly odorants (P=0.214, Mann-Whitney U-test). The mean I score on the YOF test for the normosmia group was 11.1±1.7 on a 12-point scale, whereas that for the KVSS-II was 12.1±2.3 on a 16-point scale (Table 4). For the identification test, the YOF test (mean, 0.92; SEM, 0.01) showed a significantly higher mean score ratio than the KVSS-II (mean, 0.76; SEM, 0.01; P<0.001) (Supplementary Fig. 1).

Total TDI score

The total TDI score is the sum of the scores of threshold, discrimination, and identification subtests. The mean total TDI score of the YOF test was 24.2±4.5 on a 36-point scale, whereas the mean total TDI score of the KVSS-II was 30.4±5.9 on a 48-point scale (Table 4). There was a significant difference in the mean score ratio of the total TDI score between the YOF test (mean, 0.67; SEM, 0.01) and the KVSS-II (mean, 0.63; SEM, 0.01; P<0.001) (Supplementary Fig. 1).

Optimized diagnostic cutoff of the YOF test and comparison with the KVSS-II

Fig. 3 shows the ROC curve of the YOF test and the KVSS-II for anosmia and hyposmia.

Fig. 3.

The receiver operating characteristic (ROC) curve for the total TDI scores of the YSK olfactory function (YOF) test and Korean version of the Sniffin’ stick test (KVSS-II) to predict anosmia (A) and hyposmia (B). The diagnostic cutoffs of the YOF test calculated through the Youden index were as follows: anosmia (TDI score ≤14.5: AUC, 0.88; P<0.001; Youden J score: 0.67; sensitivity, 79.8%; specificity, 87.2%) and hyposmia (14.5<TDI score ≤21.0; AUC, 0.72; P<0.001; Youden J score, 0.38; sensitivity, 55.5%; specificity, 82.1%). The diagnostic cutoffs of the KVSS-II calculated through the Youden index were as follows: anosmia (TDI score ≤20.0; AUC, 0.88; P<0.001; Youden J score, 0.67; sensitivity, 85.1%; specificity, 81.4%) and hyposmia (20.0<TDI score≤25.5; AUC, 0.72; P<0.001, Youden J score, 0.33; sensitivity, 51.1%; specificity, 83.0%). There was no significant difference in the AUC for anosmia (P=0.843) and hyposmia (P=0.902) between the YOF test and the KVSS-II. Points corresponding to the Youden J score were marked as circles on the ROC curve. T, threshold; D, discrimination; I, identification; AUC, area under the curve.

Cutoff of total TDI score for anosmia

The AUC of the ROC curve of the total TDI score of the YOF test for anosmia was 0.88 (95% confidence interval [CI], 0.85–0.89) (Fig. 3A). The AUC and the Youden J score (0.67) were maximized when the cutoff point was set to a TDI score of 14.5. Sensitivity and specificity were 79.8% and 87.2%, respectively. There was no significant difference in the AUC of the ROC curve for anosmia between the YOF test and the KVSS-II (AUC, 0.88; 95% CI, 0.86–0.90; P=0.843).

Cutoff of total TDI score for hyposmia

The AUC of the ROC curve of the total TDI score of the YOF test for hyposmia was 0.72 (95% CI, 0.69–0.75) (Fig. 3B). The optimized cutoff of the TDI score of YOF test for hyposmia was 21.0, and the Youden J score was 0.38. Sensitivity and specificity were 55.5% and 82.1%, respectively. There was also no significant difference in the AUC of the ROC curve for anosmia between the YOF test and the KVSS-II (AUC, 0.72; 95% CI, 0.69–0.75; P=0.902).

Correlations between the YOF test and the KVSS-II

We found positive correlations for each T/D/I subtest and the total TDI scores between the YOF test and the KVSS-II (Fig. 4). The Pearson correlation coefficient (r) between the YOF test and the KVSS-II was r=0.57 for the T test, r=0.65 for the D test, r=0.80 for the I test, and r=0.86 for the total TDI score (P<0.001, respectively). The total TDI scores showed a higher correlation between the YOF test and the KVSS-II than any of the T/D/I subtest scores.

Fig. 4.

Correlations between the YSK olfactory function (YOF) test and the Korean version of Sniffin’ stick test (KVSS-II). (A) Threshold test, (B) discrimination test, (C) identification test, (D) total TDI score. The correlation coefficients for the YOF test and KVSS-II were 0.57 for T, 0.65 for D, 0.80 for I, and 0.86 for the TDI score (P<0.001, respectively). T, threshold; D, discrimination; I, identification.

DISCUSSION

We developed a new olfactory test suitable for evaluating olfactory function in Koreans. Furthermore, the odorants of the identification test of the YOF test were constructed considering their major chemical functional groups. The test adopted a 12-point scale for threshold detection, discrimination, and odor identification, similar to other modern psychophysical olfactory function test batteries [8]. According to our results, the YOF test was closely correlated with the KVSS-II and also showed good performance in distinguishing Koreans’ olfactory function into normosmia, hyposmia, and anosmia.

The most commonly used odorant in threshold tests is n-butanol, which has been adopted in many olfactory tests such as the Sniffin’ stick test, the Connecticut Chemosensory Clinical Research Center olfactory test, and butanol threshold tests. However, as seen in the results of several animal experiments, inhalation of n-butanol may lead to neurotoxicity and developmental risks at sufficient concentrations [18-20]. The potential inhalation toxicity of n-butanol has been a focus of recent attention [21,22], indicating a need for an alternative olfactory threshold test to replace n-butanol.

PEA is another widely used odorant in olfactory threshold tests. PEA has several characteristics that are suitable for an odorant for olfactory threshold testing. First of all, it is a safe chemical, with minimal inhalation toxicity [23], and it has long been used as a raw material for flower-scented cosmetics. Secondly, PEA stimulates the trigeminal nerve less than n-butanol [9]. The simultaneous stimulation of the trigeminal nerve along with the olfactory nerve in olfactory function test is a longstanding issue in the selection of odorants for olfactory examinations. Trigeminal stimulation can affect the detection threshold of chemicals [24], so minimizing its effect is important in establishing a reliable olfactory function test. Finally, the simple odorhedonic characteristic of PEA was considered in the choice of an odorant for the threshold test. For example, some people have good feelings about a certain odor, while others hate the same scent. If an odor has these complex hedonic properties, it may affect the olfactory test results. Pleasant and unpleasant odors are processed by different brain regions in the orbitofrontal cortex [25], causing different autonomic responses [26]. Moreover, the detection concentration of an odorant may vary depending on the environment of each individual. For example, the threshold for food flavor can be different in states of satiety and hunger [27]. Therefore, it can be advantageous for an odorant in an olfactory threshold test to have simple hedonic characteristics. Odor hedonics are not necessarily proportional to intensity. However, the odor-hedonic characteristics of PEA (pleasantness) and n-butanol (unpleasantness)—the two most representative odorants used in threshold tests—are simply proportional to their intensity [28]. Moreover, odor hedonics can change with concentration and intensity. Even n-butanol, which is considered to have neutral characteristics from the perspective of odor hedonics, has a change in its hedonic characteristics according to its concentration [29], whereas PEA is simply pleasant regardless of its concentration. These simple hedonic features of PEA make the results of the threshold test using PEA simple to interpret.

Comparisons of n-butanol and PEA thresholds have reported contradictory results. Croy et al. [30] reported no significant difference in thresholds between n-butanol and PEA. However, Zernecke et al. [31,32] reported conflicting results in two subsequent studies. Their findings suggested that the PEA (highest concentration, 4%; 1:2 dilution ratio) threshold scores were significantly superior to the n-butanol (highest concentration, 4%; 1:2 dilution ratio) threshold scores.

We initially considered a PEA threshold test with the highest concentration at 4%, like the Sniffin’ stick test. However, in the results of a preliminary study (highest concentration, 4%; 1:2 dilution ratio; 12 steps), the olfactory threshold distribution of normosmic subjects was much lower than expected (data not shown). After adjusting the highest concentration to 10%, the PEA threshold test produced a roughly normal distribution at 5–6 points, the midpoint of the 12 steps in the test (Fig. 2A). The detection threshold of n-butanol and PEA differs from study to study for several reasons. First, depending on the solvent, the detection threshold may vary, even for the same odorant [33]. The threshold test of the Sniffin’ stick test, which uses n-butanol, or PEA threshold tests incorporate 16 steps of 1:2 dilutions with propylene glycol solvent, whereas the YOF test utilizes DPG, which is widely used as a solvent for cosmetics. This difference may cause a difference in the detection threshold concentration of PEA. Secondly, it is possible that the study populations differed across studies because there are no objective gold-standard criteria for normal olfactory function.

In the study, comparing the threshold test results with the 16-step wide step method (i.e., the method utilized in the Sniffin’ stick test) and the eight-step narrow step method demonstrated that the wide step and narrow step showed significantly reliable results, while the narrow step method showed a significantly shorter test time than the wide step method [30]. Another study investigated the reliability of reducing the number of stimuli in the threshold test using n-butanol. According to that report, using 12 steps for the threshold test had sufficient reliability, while saving test time compared to the 16-step method [34].

The KVSS-II contains odorants familiar to Koreans; however, 20 years after its development, cultural experiences have changed among generations, meaning that some odorants have become unfamiliar to Koreans. A recent study of the KVSS-II, in which the distractors were modified, reported that the correct identification percentages for normal olfactory function subjects (n=83) ranged from 36% to 96% [35]. Five of the 16 odorants (leather, lemon, licorice, resin, and apple) showed a correct answer rate less than 70%. These data are similar to our KVSS-II results (the percentage of correct answers ranged from 32.6% to 98.6%), and four of the 16 odorants had a lower correct answer rate than 70% (data not shown). The odor identification task in the YOF test was designed to use Korean culture-friendly odorants. The percentage of correct answers for each odor was relatively evenly distributed (mean, 93.6%; SD, 5.1%; range, 80.7%–98.6%). According to our results, the score ratio of the identification test of the YOF test in the normosmia group was superior to that of the KVSS-II (Supplementary Fig. 1).

Generally, the molecular acceptance range of an olfactory receptor is determined by the chemical functional group. Olfactory receptors are sensitive to aldehydes or ketones, aromatic rings, or chain lengths, and so on [36]. Different odors are recognized by different combinations of olfactory receptors [37]. Therefore, if the olfactory function test can selectively provide detailed information about the major functional groups of odorants, it would be possible to interpret the results in a more complex and elaborate manner and to use them clinically. For example, if an Alzheimer disease patient shows a pattern of impaired olfactory function for a particular odorant, it may be possible that sensitivity to the major functional group of that odorant can be used to predict the likelihood of developing the disease (i.e., Alzheimer disease).

The diagnostic cutoffs for anosmia (TDI score ≤14.5) and hyposmia (14.5< TDI score ≤21.0) were obtained by calculating the Youden index, in which the maximum point indicates the point with the maximum sum of sensitivity and specificity. The Youden index of the YOF test was much higher for anosmia (0.67) than for hyposmia (0.38), suggesting that anosmia and hyposmia are more clearly distinguished by the YOF test than hyposmia and normosmia. Based on a comparison of the AUC values of the YOF test to those of the KVSS-II, the YOF test showed equivalent results to the KVSS-II in distinguishing anosmia and hyposmia.

A comparison of the YOF test and KVSS-II demonstrated that the total TDI score and the respective subtest scores for threshold, discrimination, and identification were closely correlated. The Pearson correlation coefficient of the total TDI scores between the YOF test and KVSS-II was higher than that of each T/D/I subtest. Furthermore, on the YOF test for anosmia determination, the AUC value of the total TDI score was higher than that of each T/D/I subtest (data not shown). As described above, threshold detection, discrimination, and odor information might be affected by different levels of cognition/memory [6]. Taking these results into account, analyzing the total TDI score can provide more comprehensive and accurate information about individual olfactory function than analyzing each T/D/I subtest separately [38].

In conclusion, we developed an olfactory test that uses safe and Korean culture-friendly odorants. Various olfactory characteristics were reliably evaluated using the YOF test and demonstrated validity in comparison to another widely used olfactory function test. The YOF test is expected to provide a sufficiently comprehensive olfactory assessment that saves time compared to conventional tests. Thus, we could conclude that the YOF test proved useful to identify the olfactory function of Koreans. Further studies are being conducted to identify the safety of the odorants of the YOF test and to determine whether the sensitivity of major functional groups varies depending on the etiology of olfactory loss.

HIGHLIGHTS

Supplementary Materials

Article information Continued