Materials and Methods
Study Design and Population This prospective cross-sectional study
involved adult patients presenting to participating otolaryngology
clinics in Canada, Italy, and Spain, from May to August 2021. STROBE
Statement was used as a guideline for reporting the results.
All adults (at least 18 years old) presenting to participating
otolaryngology clinics with a pre-existing need for nasopharyngoscopy
were screened for inclusion in the study. Exclusion criteria included an
inability to understand the consent process, severe nasal obstruction at
clinician’s discretion, previous facial cosmetic surgery, history of
middle facial third orthognathic surgery or fractures, previous head and
neck oncologic resection, or facial reconstruction altering measurement
sites. Participants provided written consent before entering the study.
This study was performed in two steps. The primary objective was to
gather normative data on nasopharyngeal depth in adults and assess its
correlation with external facial measurements. Subsequently, we
developed and validated a clinical predictor to aid in nasopharyngeal
swab test performance.
Assessments
Self-reported participant demographics were collected, including age,
sex, and ethnicity. The availability of a previous computed tomography
(CT) scan of the sinonasal region was noted if present.
External facial measurements
Using a flexible surgical ruler, two external facial measurements were
assessed in cm: the distance along the curvature of the face from the
tragus to the alar-facial groove, and the distance anteriorly from the
tragus to a plane perpendicular to the philtrum (Figure 1-A and 1-B).
These variables are henceforth referred to as curved and perpendicular
distances, respectively.
Nasopharynx depth (Nasoendoscopy)
Direct measurement of nasopharynx depth (ND) was performed using a
flexible video endoscope by carefully bringing it into contact with the
posterior wall of the nasopharynx along the floor of the nasal cavity.
The point of the endoscope at the nasal sill was noted, and the
instrument was removed. ND was determined by measuring the distance from
the marked point to the tip of the endoscope in cm (Figure 2).
CT scan measurement
When a sinus CT scan was available, the distance in cm from the
nasopharynx to the nasal sill was assessed along the floor of the nose
in the axial plane.
Statistical analysis
Descriptive statistics were expressed in mean and standard deviation for
continuous variables where normal distribution was assumed, whereas
count and percentages were used to present the categorical ones.
Statistical analysis was conducted using the program SPSS for Windows
(version 20.0, Chicago, IL).
Step 1
All four measurements were presented as mean and standard deviation for
the total group, and again for each sex. An independent samples T-test
was used to compare groups, reported as a difference of mean, 95% CI
and p-value. An analysis of variance (ANOVA) was performed for each sex
to compare variables between ethnic groups.
Considering ND as the gold standard in the study, curved and
perpendicular distances were compared against the ND by subtracting each
participant’s measurement from their ND. The resulting variables
represented an error estimation between each facial measurement and ND
(i.e., Δ Perpendicular = Perpendicular - ND; Δ Curved = Curved - ND; Δ
CT scan = CT scan - ND). Thus, negative error estimation values show an
underestimation of ND, whereas positive values express ND
overestimation.
In addition, we analyzed the correlation between ND, facial
measurements, CT distance, ethnicity, sex, and age using the Pearson
correlation analysis (r).
Step 2
The second step investigated the extent to which each clinical variable
could function as a predictor for nasopharyngeal depth. A random number
sequence generator was used to randomly select a group of 271
participants. Their data was input into multiple linear regression
analyses to determine whether the external facial distances
significantly predicted ND while controlling for sex and age. Then, the
generated regression equations were validated using the remaining 100
study participants data by calculating the prediction error compared to
the actual assessed ND for each participant (error = actual –
predicted). The regression models’ errors were presented as mean,
standard deviation and 95%CI.