This cutaway view of the right nasal cavity shows the subdivided nasal regions (compartments) through a transparent septum. See "Volume Measurements" subsection of the "Subjects and Methods" section for a full description of the anatomical landmarks used to define the boundaries of each nasal compartment.
Hornung DE, Leopold DA. Relationship Between Uninasal Anatomy and Uninasal Olfactory Ability. Arch Otolaryngol Head Neck Surg. 1999;125(1):53-58. doi:10.1001/archotol.125.1.53
To examine the relationship between uninasal anatomy and olfactory ability.
A stepwise analysis of variance was used to regress the logarithm of the percentage of correct responses on the Odorant Confusion Matrix (a measure of olfactory ability) against the logarithm of nasal volume measurements determined from computed tomographic scans.
Nineteen patients with hyposmia whose olfactory losses were thought to be related to conductive disorders.
After correcting for sex differences, a mathematical model was developed in which the volume of 6 regions of the nasal cavity, 6 first-order interactions, and 3 second-order interactions accounted for 97% of the variation in the measure of olfactory ability.
Increases in the size of compartments of the nasal cavity around the olfactory cleft generally increase olfactory ability. Also, anatomical differences in the nasal cavities of men and women may account, in part, for sex differences in olfactory ability.
THE REGIONAL anatomy of the nasal cavity determines the course that incoming air takes as it moves toward the lungs. As nasal anatomy affects the pattern of airflow through the nose, it in turn affects the access of odorant molecules to the olfactory cleft (the head space above the olfactory receptors). Although not yet described quantitatively, a relationship may exist between nasal anatomy and olfactory ability. As an example, consider the improvement in olfactory ability that sometimes results in patients who have had nasal polyps removed.
Schneider and Wolf1 used 656 rhinoscopic measurements to assess the degree of nasal obstruction in 8 patients. By correlating the degree of nasal obstruction to olfactory thresholds, these investigators concluded that, as the nose becomes more obstructed, olfactory ability decreases.
Rous and Kober2 studied the relationship between nasal anatomy and olfactory ability in 50 patients. Again, using rhinomanometry and olfactory threshold measurements, the authors concluded that, after a threshold level of obstruction is reached, uninasal narrowing results in an elevated olfactory threshold. In an epidemiological study, Zusho and Okamoto3 suggested that nasal obstruction was the likely cause of the olfactory loss in about half of the 1376 dysosmic patients they studied.
Vainio-Mattila4 was the first to attempt to describe the relationship between the size of a specific nasal area and olfactory ability. He observed that the olfactory ability was better when either the nostril shape was wider near the nasal tip or the area just posterior to the nasal valve was larger. In addition, the greater a person's nasal septal deformity is, the poorer is that person's olfactory ability. Because Vainio-Mattila did not find a correlation between nasal pressures and the degree of septal deformity, he suggested that the deformity reduced olfactory ability by altering airflow through the nose.
Another way of studying how structure affects function is to determine how nasal surgery affects olfactory ability. Following this approach, Goldwyn and Shore5 and Kittel and Waller6 found improvement in the olfactory ability in some of their patients after surgical treatment, with those who had extreme septal deviations before surgery being the most likely to improve.5
Nasal modeling provides another avenue for studying the relationship between nasal anatomy and olfactory ability. Scherer and coworkers7- 10 have completed studies using engineering and modeling methods to describe the conductive mechanisms that determine the access of odorant molecules to the olfactory receptor cells. After building a model of the human nasal cavity that was 25 times larger than the actual nose, these investigators used hot-film anemometers to measure airflow velocities through the model. They have observed that in a healthy nose, about 10% of the total inspired air passes through the olfactory cleft. These investigators further observed, confirming the earlier work of Masing,11 that all the air reaching the olfactory region comes from a stream located on the ventral tip of the nares.
Leopold12 studied the relationship between nasal cavity anatomy and olfactory ability by correlating measurements from computed tomographic (CT) scans of the nasal cavity with the results of a measure of bilateral olfactory ability. Leopold concluded that changes in the volumes of the nasal cavities immediately above the middle turbinates were responsible for the greatest changes in binasal olfactory ability.
Although Leopold's study was a reasonable first step, a full understanding of the relationship between nasal volume and olfactory ability is unlikely to be achieved by combining the volumes of nasal cavities and comparing these with binasal olfactory ability. As an extreme example, if in a particular patient the olfactory ability were confined to the right nostril, unless it dramatically affected the airflow through the right nostril, changes in the structure of the left nostril would likely have no effect on the binasal olfactory ability. Indeed, in this example, adding the volumes of the 2 nasal cavities together may obscure important relationships between uninasal anatomy and olfactory ability. Studying each nostril individually seems even more compelling with the recent finding13 that in patients, bilateral olfactory ability usually reflects the activity of the better nostril.
Thus, the present study was undertaken to quantify the relationship between the anatomy of individual nostrils and their uninasal olfactory ability. A more complete understanding of how nasal anatomy affects olfactory function might have significance in the diagnosis and treatment of olfactory dysfunctions. If certain nasal areas (or combinations of areas) are found to have special importance in maintaining olfactory function, nasal surgeons would want to make every effort to repair or preserve these regions. In addition, we hope that the results of this study will aid the evaluation of the effect that treatments designed to reduce nasal obstructions would have on olfactory function.
The methods followed were similar to those described by Leopold,12 except that uninasal olfactory ability was compared with the anatomy of the nasal cavity of individual nostrils. The volumes of specific regions of the nasal passages were determined from a CT scan of nasal cavities. These volumes were then used to identify statistically the regions in the nose that best accounted for the variation in the olfactory ability. This study was designed to generate a regression model that might help explain the relationship between nasal anatomy and olfactory ability. We developed this model with the hope that it would generate testable hypotheses about how the sense of smell is related to the structure of the nose.
Subjects for this study were patients with olfactory problems that were evaluated at the State University of New York–Syracuse Health Science Center Smell and Taste Disorders Clinic. This evaluation included a history, physical examination of the head and neck, olfactory testing, and a full CT scan of the cranium and nasal cavities. Because the goal of this study was to examine the relationship between olfactory ability and nasal anatomy, patients whose olfactory losses were not thought to be related to a conduction (access) problem were eliminated from the study. This was done because nasal anatomy was unlikely to significantly influence olfactory ability in patients whose smell losses were due to problems that were primarily neural (ie, head trauma, brain tumors, etc). Likewise, patients who reported dramatic fluctuations in their sense of smell were not included. Patients who were considered to have anosmia or normosmia were also eliminated because anosmic patients were unlikely to show a relationship between nasal anatomy and olfactory ability. Normosmic patients were excluded because some upper limit to the relationship between anatomy and olfactory ability was expected. That is, if a patient could correctly identify all the odors in the olfactory test, increasing the size of nasal compartments could not, because of the ceiling of the test, increase the reported olfactory ability. Because the olfactory ability was evaluated uninasally, a few patients who had hyposmia in 1 nostril were found to have anosmia in the other nostril. For these patients, only data from the hyposmic nostril were used. The causes of the hyposmia observed in the 19 patients included in this study were, in most cases, likely related to nasal polyposis or mucosal edema. All subjects provided written informed consent under a protocol approved by the Institutional Review Board of the Public Health Service.
Each subject's uninasal olfactory ability was evaluated using the Odorant Confusion Matrix (OCM).14 Within a single block, subjects were presented 10 odorants, 1 at a time, in randomized order and asked to identify the name of the odorant from a list of 10 common substances. After 10 such blocks (100 trials), the percentage of correct responses was calculated. This is a ratio of the number of correct identifications divided by the number of trials. Because of random fluctuations in the number of correct guesses in 100 trials, subjects who score 20% or less are considered to have anosmia. Based on test results from subjects who have a normal sense of smell, a score of 80% or better on the OCM is considered to reflect normosmia. A full (100-trial) OCM was conducted for each nostril.
The olfactory testing was performed within 3 days of the CT scan, and in almost all subjects (15 of 19), the measure of olfactory ability and the nasal imaging were completed on the same day.
Horizontal CT scans were performed on a CT scanner (model 9800, General Electric Corp, Milwaukee, Wis). Because the goal of this study was to describe the relationship between the volume of specific nasal areas and olfactory ability, the nasal passages were subdivided into 21 different compartments, which are represented diagrammatically in Figure 1. Because the cribriform plate was recognizable on every CT scan, its anterior and posterior borders were used to divide the nasal cavity coronally into 3 regions: anterior to the cribriform, below the cribriform, and posterior to the cribriform. These divisions extended from the cribriform to the roof of the palate. The regions above the inferior margin of the middle turbinate were further divided horizontally into the top, middle, and bottom 5-mm slices below the cribriform. These regions (numbers 1-9) are located between the middle turbinate and the septum. The regions below the margin of the middle turbinate (bottom of the third 5-mm slice) and the palate were divided horizontally by a plane parallel to the palate and halfway between the palate and the middle turbinate. The lateral boundary of these compartments was a plane from the margin of the middle turbinate to the palate, parallel to the septum. These compartments (numbers 10-15) were located between the middle turbinate and the roof of the mouth. The lateral meatus were divided by a plane drawn from the middle of the cribriform plate perpendicular to the septum. The bottom of the third 5-mm slice was extended to divide the upper lateral meatus. These regions were designated by the numbers 16 through 21.
The density range of the CT scan was set so that air could be easily separated from tissue. After the boundaries of each compartment were delineated and the airspace densities identified, the software system of the CT scanner calculated the area (in square centimeters). The thickness of the slice provided the third dimension necessary for a volume determination. The slice-by-slice volumes were added to determine the entire air volume for each compartment
Because humans have a complete nasal septum, because with odorant identification tests the 2 nostrils have been found to make independent contributions to olfactory ability,13 and because the olfactory ability was measured uninasally, for a given subject, data from both nostrils were included when each nostril met the selection criteria described earlier. For the purposes of this analysis, the data from the 2 nostrils were treated as independent variables.
We tested the hypothesis of the independence of the 2 nostrils as follows: Uninasal responses from the same subject seemingly would produce errors that were not independent of each other, thus violating an assumption that is fundamental to multiple linear regression models. As uninasal anatomical covariates were selected for inclusion in the final model, however, possible sources of dependency between regressor model errors from the same subject were expected to be progressively removed. Whether the selected uninasal covariates were completely successful in producing a regression model with independent errors was an issue that still needed to be addressed. For this purpose, a 1-way analysis of variance was performed on the pairs of errors in the final model for the subjects with both left and right uninasal scores. A 2-tailed variance ratio (F ratio) was used to test the fundamental assumption of independent errors. A nonsignificant F ratio would validate the use of each nostril as an analytic unit in the final regression model. A significant F ratio would indicate the need for additional model development before accepting the assumption of independent errors. If the test of the assumption of the independence of the 2 nostrils demonstrates that the 2 nostrils do not behave independently, considerable care would need to be exercised in interpreting the results of the final model.
Because Leopold12 had previously shown that the natural logarithm of the percentage of correct responses on the OCM vs the natural logarithm of the independent variables was an appropriate manner with which to account for the variations in the OCM performances, we developed this model for the present study. The use of the logarithm transformation seemed appropriate because, like most sensory systems, olfactory information is generally encoded as a logarithm function. In addition, because relatively small changes in the nasal anatomy usually do not produce dramatic changes in olfactory function, we hypothesized that a logarithm transformation would more suitably describe the anatomical conditions that would dramatically affect olfactory function.
To eliminate sources of variability beyond those due to volume changes, nonvolume variables were first considered as candidate variables. Nonvolume variables included the nostril (left or right), the patient number, the angle of the CT cut, the diagnostic group, the presence of nasal disease, the number of nostrils contributed to the study (1 or 2), and the patient's age and sex. Volume variables were determined by adding compartments 1, 2, and 3; 4, 5, and 6; 7, 8, and 9; 1, 4, and 7; 2, 5, and 8; 3, 6, and 9, 11 and 12; 14 and 15; 10 and 13; 16 and 17; 18 and 19; 20 and 21; and the sum of all 21 compartments. In addition, ratios of the following volumes of compartments were used as variables: 1:4, 2:5, and 3:6.
Using the natural logarithm of the uninasal OCM score as a measure of olfactory ability, a stepwise linear regression was performed to generate a mathematical model by which changes in the natural logarithm of the nonvolume variables would reflect changes in olfactory ability. A minimum ratio (F=4.00; P<.05) was used to allow variables that were to be entered in the model 1 at a time. At each step of the analysis, the next independent variable was included in the model only if an F ratio to evaluate its partial association with the number correct met the criterion (F>4.00). If the partial F ratio for any previously included variable fell below 4.00, it was dropped from the model. The stepwise up-and-down process was continued until no remaining candidate variables met the criterion for inclusion.
Candidate variables were considered in the following order: (1) nonvolume variables (medical record or data recording), (2) primary variables (compartment volumes), (3) first-order interactions (products of all included variables from stages 1 and 2), and (4) second-order interactions (products of first-order interactions and included variables from stages 1 and 2).
Data are given as the mean percentage of correct scores.
A profile of the 19 patients included in this study is shown in Table 1. For the 19 subjects (34 nostrils) enrolled in this study, the OCM score ranged from 46% to 79%, with the mean score being 62%. Although data from both nostrils were included for most subjects, single-nostril data from 2 men and 2 women did not meet the selection criterion for inclusion in the study. The OCM data shown in Table 1 represent the range for the nostrils used. The OCM scores for the single-nostril subjects were 58% and 64% for the 2 men and 46% and 51% for the 2 women. The likely causes for the smell loss are also included in Table 1. For the subjects listed in the idiopathic diagnostic group, sequelae of head trauma and upper respiratory tract infection were thought not to contribute to their olfactory loss.
Of the possible nonvolume variables, only sex met the criterion for inclusion (F=4.00; 6% of the variation accounted for). As a result, sex was always the first variable used in subsequent analyses.
Six additional variables met the criterion for inclusion as primary variables for the stepwise regression analysis (Table 2). Note that in Table 2, the final primary variable to enter the model had a partial F value of greater than 11 (P<.001). One interpretation of this high partial F value is that the likelihood of a type I error due to the high number of analyses performed is low. Another 6 variables met the criterion for inclusion in the first-order interaction (Table 3), and 3 more variables were included in the second-order interaction model (Table 4). The final model accounts for approximately 97% of the variation in the OCM score.
The hypothesis of the independence of the 2 nostrils was tested as follows: After the final model had been developed, the analysis of variance of the error terms (difference between calculated and predicted values) by nostril for the 15 subjects who contributed 2 nostrils to the study was completed. This was a 2-tailed variance ratio (F14,15=1.40; P>.46). A nonsignificant F ratio validates the use of 2 nostrils as independent analytic units in the final regression model.
The present study was designed to generate a regression model that might help explain the relationship between nasal anatomy and olfactory ability. In the generation of the final model, several candidate variables were considered for inclusion in the 2-stage analysis used in this study. Because data were from only 34 nostrils, the ratio of the number of candidate variables to the number of test sessions might be less than optimal. Given the size of the P values for the successful candidate variables, however, the present study provides a reasonable indication of the candidate variables that might help explain the relationship between olfactory ability and nasal anatomy. As we describe later, the final model allows for the generation of several testable hypotheses about how the sense of smell is related to the structure of the nose.
Sex was the first variable to be entered in the model. For the purposes of this study, 1 indicates female and 2 indicates male. Given this designation, a negative coefficient means that the olfactory ability of the men was, on average, less than the olfactory ability of the women. The observation that women have a better sense of smell than men is consistent with previously published work.15- 18
If the sex variable had entered the model without any interactions, the present study results would have been consistent with what has been previously reported.15- 18 The stepwise regression, however, revealed an interaction between sex and the volumes of the lateral meatus (the sum of the volumes of compartments 16 and 17 and the volume of compartment 20). The second-order interaction of sex with the volume of compartment 20 and several other variables also met the criterion. That both first- and second-order interactions with compartment volumes and sex were included in the final model might at least partially explain why women have a better sense of smell than men. That is, these first- and second-order interactions suggest the possibility of anatomical differences in nasal structure between women and men that contribute to their different olfactory abilities. For example, because men, on average, have larger noses, their olfactory ability might be more sensitive to changes in the size of the lateral meatus. In other words, a larger lateral meatus might shunt more of the inspired air away from the olfactory mucosa and so slightly reduce the overall olfactory ability. As a corollary, the average olfactory ability in women would be slightly better than that in men. Of course, other explanations for the sex difference in olfactory ability are possible, including differences in mucosal thickness, hormone levels, or both. Nevertheless, the interactions between sex and the size of certain nasal compartments suggest several testable hypotheses that it is hoped will be the subject of future investigations.
The first 2 primary volume variables to be entered in the model were the ratio of compartments 2 to 5 and the volume of compartment 4. Because the coefficients for these primary variables showed a positive correlation, olfactory ability increased as the size of these variables increased. For compartment 4, it may simply be that as the size of the anterior area of the olfactory cleft increases, so does olfactory ability. The relationship between olfactory ability and the ratio of compartments 2 to 5 is slightly more complex because the ratio could increase in 1 of 2 ways: The volume of compartment 2 could increase in size, or the volume of compartment 5 could decrease in size. Either of these changes would likely increase the amount of air flowing to the head space above the olfactory receptors, and this may explain the increased olfactory ability.
The model suggests that there is a positive relationship between the size of the posterior portion of the upper half of the main channel (compartment 12) and olfactory ability. In this case, a larger cavity might contribute to the laminar flow that seems to be important in maintaining olfactory ability.
As the size of the compartment at the nostril (the sum of the volumes of compartments 10 and 13) increased, the olfactory ability decreased. In this case, a smaller lateral entrance may somehow channel air toward the olfactory region. Likewise, the relationship between the sum of the volumes of compartments 16 and 17 and olfactory ability might be related to a channeling of air away from the olfactory area. In other words, as the size of the lateral meatus increases, more air is shunted away from the olfactory area, and the sense of smell is diminished. These observations seem to be consistent with the previously described conclusions of Schneider and Wolf.1
Because the volumes of the various nasal compartment are small, and because all nasal compartments are touched by at least 5 other compartments, it is not surprising that several first- and second-order interactions were observed. The possible significance of the interaction between sex and the size of some of the nasal compartments has been discussed earlier. The interactions between the volumes of some of the nasal compartments (20 and 12, 4 and 12, and the ratio of compartments 2 to 5 and compartment 20; steps 8, 11, and 13) suggest that as the size of the area of the nasal cavity below the middle turbinate increases relative to the size of some of the compartments above the middle turbinate, olfactory ability is decreased. This implies that as a larger proportion of inspired air is directed away from the olfactory area, olfactory ability is decreased. Following the same logic, one interpretation of the interaction between the volumes of the nasal compartment above the middle turbinate (ratio of 2 to 5 and compartment 4; step 9) might be that as the size of the nasal cavity anterior to the olfactory region increases relative to the ratio of the nasal volumes immediately below the olfactory area, more air is directed toward the receptors, and olfactory ability is improved. The inclusion of the first- and second-order interactions suggests that the relationship between olfactory ability and nasal structure is complex and that changing a structure in 1 part of the nose far removed from the olfactory area can have dramatic effects on olfactory ability.
Caution must be exercised in interpreting the results of the present study. The subjects chosen for this study all had a reduced sense of smell that, in most subjects, was thought to be related to a conduction problem. Therefore, at least for the present, the model is appropriate only for patients who fit this profile. How the size of various nasal compartments is related, if at all, to olfactory ability in normosmic subjects is not yet known.
Nevertheless, the ability to specify the relationship between nasal anatomy and olfactory ability should allow nasal surgeons to better evaluate how they approach their surgical patients. Furthermore, the present study should allow nasal surgeons to better evaluate whether a reduced olfactory ability is related to an anatomical configuration that affects olfactory airflow. For some patients, it might even be possible to advise them about how their nasal surgery might affect their postsurgical olfactory ability.
Accepted for publication September 9, 1998.
This work was supported in part by grants DC00220 and DC00072 from the National Institutes of Health, Bethesda, Md.
Corresponding author: David E. Hornung, PhD, Physiology Department, State University of New York–Syracuse Health Science Center, 750 E Adams St, Syracuse, NY 13210 (e-mail: Hornung@vm.stlawu.edu).