Correlation analysis between odor identification scores obtained with 3–alternative forced choice (AFC) and 6-AFC tests for both patients and healthy control individuals.
Odor identification scores obtained with the 3–alternative forced choice (AFC) and 6-AFC tests shown separately for patients and healthy control individuals. The box-whisker plots show the 5th, 25th, 50th, 75th, and 95th percentiles.
Scores for 3–alternative forced choice (AFC) and 6-AFC identification tests for subgroups with postviral and sinonasal disease and who had experienced trauma. The graph shows a significantly greater discrepancy between 3-AFC and 6-AFC test scores for the subgroup that had experienced trauma compared with the subgroup with postviral and sinonasal disease. Error bars indicate SEM.
Negoias S, Troeger C, Rombaux P, Halewyck S, Hummel T. Number of Descriptors in Cued Odor Identification Tests. Arch Otolaryngol Head Neck Surg. 2010;136(3):296-300. doi:10.1001/archoto.2009.231
To investigate whether results from cued odor identification tests are influenced by the number of verbal descriptors presented with each odor and whether a larger number of descriptors would improve the differentiation between individuals with various degrees of olfactory loss.
Randomized, crossover study.
University Clinic in Brussels.
Two hundred thirty-eight study participants (128 with olfactory concerns and 110 control individuals without such concerns). The study took place from March 1, 2008, to April 30, 2009.
Main Outcome Measures
Results of the 3–alternative forced choice (AFC) and 6-AFC identification tests, which consisted of 32 odor items each.
A high correlation was found between the results obtained with the 2 tests. Although no sex-related differences were present, age was found to interact with performance in that lower test scores were obtained for older participants with the 6-AFC test, compared with the 3-AFC test. Of the 2 tests, the 6-AFC better separated participants with self-assessed olfactory deficits from participants without such concerns. It also discriminated more strongly between posttraumatic and postviral and sinonasal underlying causes of olfactory loss.
Cued odor identification tests with various numbers of verbal descriptors produce similar results. However, an increasing number of alternative descriptive items seem to allow for better discrimination between individuals with and without olfactory loss.
Odor detection thresholds, odor discrimination, odor identification, and odor memory capacity are among the most frequent tests used to describe olfactory performance. Most of the validated tests used at the moment are partially or fully based on evaluating these dimensions, including the University of Pennsylvania Smell Identification Test,1,2 the Sniffin’ Sticks test,3,4 the European Test of Olfactory Capabilities,5 and the Connecticut Chemosensory Clinical Research Center Test.6 Because of time constraints in clinical practice, the screening variations of these tests consist mostly of odor identification tests.7- 10
Odor identification tests are often thought to reflect higher cognitive levels of olfactory processing, apart from an obviously integrated sensory detection process.11 Nevertheless, a correlation between detection thresholds and identification scores has been demonstrated.12 Furthermore, it had been suggested that sensory functions dominate the task in a cued odor identification test.13 These premises allow the sole use of identification tests in discriminating among different levels of olfactory function.
When building an identification test that aims at high specificity and sensitivity in terms of separation between healthy and subhealthy olfactory function, several aspects must be considered. Odor identification has been shown to be highly influenced by the task the participant is performing: a free choice identification task is more difficult than a cued one14 and requires a higher degree of cognitive contribution.15 Furthermore, the results of a cued identification task are influenced by whether the cues are words or pictograms3,16 or even by the color of odorized probes.17
In cued odor identification, the choice of odor descriptors also plays a major role in terms of test outcome. For example, the use of more contrasting distracters leads to a better discrimination between patients with hyposmia and anosmia,18 which, in a clinical context, is important to the prognosis and therapy of patients.19,20 Analogous to the choice of distracters, the number of alternative answers could also be important. Most of the currently available odor identification tests use a 4-item forced choice procedure. The aim of the present study was therefore to investigate whether the results from tests of odor identification are influenced by the number of verbal descriptors that are presented with the odor. In addition, a larger number of descriptors should increase the specificity of the diagnosis. We also hypothesized that women would outperform men with increasing task difficulty because women are known to have better verbal abilities than men.13 In addition, we wanted to examine how age interferes with regard to the increasing difficulty of the odor identification task.
A total of 238 individuals participated in the study, of whom 128 were visiting St Luc Hospital, Catholic University of Louvain with olfactory dysfunction concerns (patient group; 61 men and 67 women; mean [SD] age, 52.3 [15.7] years; range, 18-80 years) and 110 were healthy individuals without olfactory concerns recruited through word of mouth or posters placed in the perimeter of the University Clinic (control group; 57 men and 53 women; mean [SD] age, 49.3 [15.1] years; range, 18-77 years). The 2 groups did not differ in age (t236 = −1.505; P = .13) or age distribution (χ2 = 0.410; P = .60), smoking habits (χ2 = 0.336; P = .65), or toxic exposure to chemicals that might interfere with olfactory function (χ2 = 1.248; P = .29). The study took place from March 1, 2008, to April 30, 2009. All participants provided written informed consent. The study was conducted according to the Declaration of Helsinki on Biomedical Research Involving Human Subjects.
A detailed medical history was taken, comprising data with regard to head trauma or infection of the upper respiratory tract, medication, surgery, chronic diseases, smoking and drinking behavior, exposure to toxic substances, and duration of olfactory problem. These data, together with the results of an otorhinolaryngologic clinical examination that included nasal endoscopy, offered information on the probable cause of olfactory dysfunction in patients (Table 1) and provided information on exclusion criteria for the healthy control group (age <18 years; history of severe neurologic or medical comorbidity; head trauma; substance abuse; history of any disease known to have an impact on the olfactory function, such as Parkinson disease and renal failure; acute or severe chronic rhinitis or sinusitis; or any other relevant otorhinolaryngologic condition that can interact with the olfactory function, such as severe septum deviation, history of trauma, and nasal polyps). Additional exclusion criteria for all participants were represented by neurologic or psychiatric conditions that could affect cognitive function status. Both groups were asked to rate their olfactory function in comparison to others on a scale of no perception, very bad, much worse, worse, somewhat worse, normal, or better.
Study participants were tested with a 3–alternative forced choice (AFC) and a 6-AFC identification test. The sequence of testing was randomized across all participants. Both tests were performed in the same session, with a 10-minute break between them. The tests were an extension of the 16-item identification subtest part of the Sniffin’ Sticks battery.21 This subtest consists of 32 odor-dispensing pens, each with a length of 14 cm and an inner diameter of 1.3 cm. The absorbent material inside the felt-tip pens is filled with liquid odorants. For odor presentation the cap was removed by the experimenter for approximately 3 seconds, and the pen's tip was placed approximately 2 cm in front of both nostrils. The interval between 2 odor presentations was approximately 30 seconds. The study participants were required to smell 32 odor-containing pens and then name the appropriate answer from a list of 3 or 6 choices (Table 2).
Results were analyzed with SPSS statistical software, version 16.0, for Windows (SPSS Inc, Chicago, Illinois). Age, sex, smoking behavior, and toxic exposure comparisons between the control and patient groups were performed using the χ2 test. Results were submitted to analyses of variance, adopting test (3-AFC and 6-AFC) as the within-subject factor; the factors group (patients vs healthy control individuals), sex, and probable diagnosis were used as between-subject factors. Degrees of freedom were adjusted according to the Greenhouse-Geisser correction. The t tests for independent samples were used for additional comparisons between groups. Correlation analyses were performed using Spearman statistics. The level of significance was .05.
A high correlation was found between the results obtained with the 2 tests (r = 0.92; P < .001, Figure 1) for both the patient (r = 0.88; P < .001) and control groups (r = 0.71; P < .001). Descriptive statistics for the 2 tests are given in Table 3, including the 10th to 90th percentiles for both groups. None of the healthy control individuals scored lower than 16 or 22 on the 6-AFC or 3-AFC test, respectively.
Predictably, we found the difference between the 3-AFC and 6-AFC test results to increase with age; there was a higher rate of incorrect answers with the 6-AFC test for older people compared with younger ones (r = 0.17; P =.007). No major effect of the factor sex was found, although, on average, women outperformed men, which was more visible in the 6-AFC task (F1,233<.01; P = .98) (Table 4).
Study participants who identified themselves as having olfactory loss performed worse on both olfactory tests than study participants who rated themselves as not having such a deficit. Relative to the 3-AFC test, however, their performance was even worse on the 6-item test (F1,237 = 4.91; P=.009) (Figure 2). Furthermore, study participants who assessed their olfactory function as very bad or had no perception performed worse on the 6-AFC test compared with the 3-AFC test than the study participants who rated their olfactory perception as being bad, worse, or a bit worse, although this finding did not reach a significant level (P = .15).
Different causes of olfactory loss were found to produce different levels of olfactory dysfunction: when analyzing the 3 most common causes of olfactory loss, patients who had experienced trauma showed significantly lower results than patients with postviral and sinonasal disease individually on both tests (Figure 3; 3-item test: P=.09 and P=.01, respectively; 6-item test: P=.02 and P=.001, respectively). For patients with postviral and sinonasal conditions, the results are the same regardless of the test used (3-item test: P=1.0; 6-item test: P=1.0)) whereas a significant interaction test by cause of disorder was found when comparing patients who had experienced trauma with patients with postviral and sinonasal disease together (F1, 110 = 4.15; P=.04 < .05, Figure 3).
The major result of the present study is that the difficulty of cued odor identification increases with the number of verbal cues. This finding can be important in both a clinical and nonclinical context.
From a clinical point of view, we presumed that a more difficult task would help to better discriminate between individuals with normosmia and patients with olfactory disorder, analogous to the findings by Gudziol and Hummel.18 In the currently investigated group of individuals lacking cognitive impairment, a strong correlation was found between the 3-AFC and 6-AFC test results, with the 6-AFC test producing overall lower results compared with the 3-AFC test. Study participants who identified themselves as having an impaired sense of smell performed worse on the 3-AFC test than individuals rating themselves as not having impairment; however, as predicted, study participants who described themselves as having olfactory impairment scored even lower on the 6-AFC test.
Inside the patient group, an increasing difference between the performances on the 3-AFC test vs the 6-AFC test was noticed with self-estimated degree of olfactory impairment. Both tests revealed 2 levels of olfactory impairment according to the cause of the disorder: on one level, the patients with postviral and sinonasal disease and on the other, the patients who had experienced trauma. The 6-AFC identification test, however, was found to discriminate better between patients who had experienced trauma and patients with postviral and sinonasal disease than the 3-AFC test.
No sex-related differences were noticed with any of the tests used, although women had slightly higher odor identification scores. Conflicting findings have been reported in the literature on this subject,22 with women typically outperforming men in cases where a significant effect of sex had been described. In return, in congruence to previous studies,13,23 both tests revealed an age-related decrease of odor identification ability, significantly more pronounced with the 6-AFC test. Further tests are needed to elucidate whether this tendency is due to the (presumably) higher cognitive demands of the 6-AFC test.
A more difficult identification task would also be important when analyzing the odor identification ability within the population with normosmia. Acquisition of meaningful information is critically dependent on the idea that the odor test does not exhibit a ceiling effect. Increasing the difficulty of the test by including more tested odorants and a higher number of distracters for each targeted item might represent a solution for this issue.
One limitation of this study was the lack of use of a reference for an independent determination of olfactory function. The 2 groups of investigated individuals were selected with regard to self-identified olfactory dysfunction. Previous work indicated that self-assessment of olfactory abilities may produce conflicting results. Specifically, one study24 reported good correlation between ratings and measures in patients with smell disorders; another study25 clearly showed that the ability to correctly estimate one's own olfactory function depends on the degree of attention toward the sense of smell. Yet other studies observed low accuracy in olfactory self-assessment in different populations (eg, patients with Parkinson disease,26 patients with Alzheimer disease,27 aging individuals,28,29 or young, healthy individuals naive to olfactory tests25). However, although the self-ratings of the study participants investigated in the present study were probably not 100% accurate, the use of the 2 olfactory tests nevertheless provided the predicted results, probably also because of the large number of individuals investigated. In turn, this finding suggests that self-ratings obtained in the current study contained some degree of meaningful information.
In conclusion, the number of items offered as verbal cues in an identification task plays a significant role in the diagnosis of olfactory dysfunction. It helps to discriminate more accurately among patients with different degrees of olfactory loss.
Correspondence: Thomas Hummel, MD, Interdisciplinary Center for Smell and Taste, Department of Otorhinolaryngology, University of Dresden Medical School, Fetscherstr 74, 01307 Dresden, Germany (firstname.lastname@example.org).
Submitted for Publication: July 15, 2009; final revision received August 11, 2009; accepted September 29, 2009.
Author Contributions: Dr Negoias had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Hummel. Acquisition of data: Troeger, Rombaux, and Halewyck. Analysis and interpretation of data: Negoias. Drafting of the manuscript: Negoias and Troeger. Critical revision of the manuscript for important intellectual content: Rombaux, Halewyck, and Hummel. Statistical analysis: Negoias and Hummel. Administrative, technical, and material support: Troeger and Hummel. Study supervision: Rombaux and Hummel.
Financial Disclosure: None reported.
Additional Contributions: Ilona Croy, Dr.rer.medic, provided helpful comments regarding the manuscript.