You are here

Article Text

Article menu
Original article
Diagnostic accuracy of lateral neck radiography in ruling out supraglottitis: a prospective observational study
  1. Takashi Fujiwara1,
  2. Hiroshi Okamoto2,
  3. Yasuhiro Ohnishi3,
  4. Toshio Fukuoka4,
  5. Kazuyuki Ichimaru5
  1. 1Department of Otolaryngology Head and Neck Surgery, Ehime University, Toon City, Ehime, Japan
  2. 2Department of Emergency Medicine, Kurashiki Central Hospital, Kurashiki City, Okayama, Japan
  3. 3Department of Radiology, Kurashiki Central Hospital, Kurashiki City, Okayama, Japan
  4. 4Department of Emergency Medicine, Kurashiki Central Hospital, Kurashiki City, Okayama, Japan
  5. 5Department of Otolaryngology Head and Neck Surgery, Kokura Memorial Hospital, Kitakyushu City, Fukuoka, Japan
  1. Correspondence to T Fujiwara, Department of Otolaryngology Head and Neck Surgery, Ehime University, Shitsukawa, Toon City, Ehime 791-0295, Japan; t.fujiwarabi{at}gmail.com

Abstract

Objective To assess the diagnostic accuracy of lateral neck radiographs (LNR) for acute supraglottitis in adults.

Design A single centre prospective observational study.

Setting Emergency department at Kurashiki Central Hospital, Japan.

Participants Adult patients who underwent LNR to detect supraglottitis.

Main outcome measures Presence of supraglottitis, based on nasopharyngeal laryngoscopy or a follow-up telephone call, 7–30 days after the visit.

Results 140 patients had LNR during the study period. 35 patients were excluded from further analysis because of lack of consent. Of the 105 eligible patients, 21 patients (20%) were given the diagnosis of supraglottitis: 17 of 29 with a radiographic abnormality, and 4 of 76 patients without a radiographic abnormality. Three of the four cases where LNR was negative was grade 1, and all cases of grade 3 or higher had abnormal LNR. Sensitivity and specificity (95% CI) of LNR for supraglottitis were 81.0% (64.2 to 97.7) and 85.7% (78.2 to 93.2), respectively. The positive predictive value of LNR was 58.6% (40.7 to 76.5) and the negative predictive value was 94.7% (89.7 to 99.8). The positive likelihood ratio of LNR was 5.67 (3.27 to 9.82) and the negative likelihood ratio was 0.22 (0.10 to 0.51).

Conclusions LNR showed only moderate sensitivity and specificity for supraglottitis and would miss some cases of supraglottitis if the pre-test probability is high. LNR was very sensitive for grade 3 or higher supraglottitis, but would miss milder cases.

Trial registration UMIN000011928.

  • Airway
  • ENT
  • Imaging, x-ray
  • Infectious Diseases, Bacterial

https://doi.org/10.1136/emermed-2013-203340

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Key messages

    • What this paper adds

    • Observational studies have suggested that lateral neck radiographs can be used to rule out supraglottitis, but the diagnostic accuracy of the test in prior studies may be affected by spectrum bias.

    • In this prospective study, lateral neck radiographs had a PPV of 58.6% and NPV of 94.5%, with a negative LR of 0.22. Radiographs could not rule out all cases of supraglottitis in this prospective study and appears to miss milder cases.

    Introduction

    Supraglottitis is an inflammation of the epiglottis and surrounding soft tissues, and can cause sudden life threatening airway obstruction which may require emergency airway management.1 The most common symptom of supraglottitis is a sore throat; however, this is a very common symptom in the emergency setting, resulting mostly from self-resolving diseases, such as the common cold. Ruling out supraglottitis is thus challenging for the emergency physician. Direct visualisation of the epiglottis by nasopharyngeal laryngoscopy or indirect laryngoscopy is the gold standard for a diagnosis of supraglottitis; however, these tests require special training and expertise.2 Soft tissue lateral neck radiographs (LNR) are a common procedure for screening for supraglottitis because they are easily obtained and available in most clinical settings.3 ,4 Abnormal findings on LNR in patients with supraglottitis are thickening of the epiglottis, called the ‘thumb sign’,5 ,6 and a poorly defined vallecula air pocket, termed the ‘vallecula sign’.7 Previous studies reported high sensitivity (83–100%) and specificity (87–100%) of LNR for supraglottitis.6–8 However, these retrospective studies calculated diagnostic accuracy using inappropriate spectrum patients. The diagnostic value of LNR was calculated using patients with supraglottitis diagnosed by laryngoscopy. Generally, LNR are used to screen for, and laryngoscopy is not always performed in patients who have normal LNR.6 ,7 Previous studies did not sufficiently address those patients who had radiographs but did not have laryngoscopies. Also, the specificity of LNR was calculated using patients with trauma and foreign bodies as the control group,7 although LNR of lingual tonsillitis patients, who present with similar symptoms to supraglottitis, may also have abnormal radiographs.9 Spectrum bias is associated with a risk of overestimating the test accuracy.10 Therefore, we conducted a prospective validation study to determine the diagnostic accuracy of LNR in the diagnosis of supraglottitis in adults.

    Methods

    Study population and protocol

    We performed a single centre, non-interventional, prospective study in the emergency department of Kurashiki Central Hospital, from 16 April 2011 to 31 March 2013. The study was conducted and reported in accordance with Standards for Reporting of Diagnostic Accuracy principles.11 We prospectively recruited consecutive adult patients (>15 years old) whom had undergone LNR to evaluate supraglottitis. LNR were performed if patients had a sore throat, odynophagia, drooling, voice change or dyspnoea. The emergency department physicians decided whether the patients needed LNR, and trained radiological technologists performed the LNR. Nasopharyngeal laryngoscopy was performed in patients with radiographic abnormalities, as well as in those without abnormalities but in whom the emergency physician still suspected supraglottitis. All enrolled patients were followedup by telephone interview after 7–30 days. Planned exclusions were patients who did not provide consent or who were not followed up by telephone interview.

    For each enrolled patient, data on patient characteristics, symptoms and the presence of supraglottitis were collected from the medical records. The institutional review board of Kurashiki Central Hospital approved the study.

    In contrast with therapeutic or interventional studies, formal sample size calculations based on power assumptions for diagnostic modelling cohort studies do not exist. As we intended to roughly calculate the sensitivity and specificity of LNR in the clinical setting, we defined the sample size as 18 patients with supraglottitis; this number is similar to the numbers in previous studies.5–7 The incidence of supraglottitis in adults is 1–3 per 100 000/year.12 ,13 Kurashiki Central Hospital is an urban hospital that serves 800 000 people in the western area of Okayama prefecture, and there are approximately 70 000 visits to the hospital emergency department annually. At Kurashiki Central Hospital, 10–15 patients with supraglottitis were hospitalised annually from 2000 to 2010, and about 20% were diagnosed without LNR. We set a 2 year study duration to enroll a sufficient number of patients with disease.

    Index and reference standard

    The index test was the LNR. A trained radiological technologist performed all LNR. One emergency physician (HO) and one radiologist (YO) independently interpreted each LNR. HO and YO were blind to the patient's history, symptoms and laryngoscopic results. Disagreement was resolved by discussion, and the presence of a ‘thumb sign’ and ‘vallecula sign’ were recorded. Interobserver differences between HO and YO were analysed using k statistics, including 95% CI.

    The reference standard was the findings of nasopharyngeal laryngoscopy or follow-up telephone interview after 7 days. The laryngoscopies were performed in the emergency department or otolaryngology department by otolaryngologists. Two otolaryngologists (TaF and KI) classified the findings according to Katori's laryngoscopic grading system: grade 1: the patient's epiglottis is slightly swollen with the entire length of the vocal folds visible by scope; grade 2: the patient's epiglottis is moderately swollen with more than half of the posterior vocal folds visible by scope; and grade 3: the patient's epiglottis is severely swollen with less than half of the posterior vocal folds visible by scope (figure 1).14 Emergency department physicians decided whether to prescribe antibiotics before or after the laryngoscopy or if the patient did not undergo the procedure. For those who did not undergo nasopharyngeal laryngoscopy during the visit, we conducted a follow-up telephone interview after 7 days because supraglottitis generally worsens within 7 days. If we did not reach participants at this time, we called repeatedly in an attempt to reach them within 30 days of the visit. If the patient's symptoms had resolved, we defined the outcome as negative for supraglottitis. If the patient visited the emergency department repeatedly, we defined the outcome as positive or negative based on the diagnosis at the subsequent emergency department visits. We also asked whether patients were prescribed antibiotics from other hospitals or physicians.

    Figure 1
    Figure 1

    Katori's scope grading of supraglottitis. (A) Grade 1: the patient's epiglottis is slightly swelled and the entire length of the vocal folds can be seen with the scope. (B) Grade 2: the patient's epiglottis is moderately swelled and more than half of the posterior vocal folds can be seen with the scope. (C) Grade 3: the patient's epiglottis is severely swelled and only less than half of the posterior vocal folds can be seen with the scope.

    Statistical analysis

    Using the laryngoscopic diagnosis of supraglottitis or the telephone interview as the standard, we calculated the test characteristics (sensitivity, specificity, predictive value and likelihood ratios) of LNR with 95% CI using the normal approximation. All statistical analyses were performed using SPSS software V.19.0 (SPSS Inc, Chicago, Illinois, USA). The first author (TaF) had full access to all study data and analyses.

    Results

    From 16 April 2011 to 31 March 2013, 140 LNR were performed to rule out supraglottitis, and 105 LNR met the inclusion and exclusion criteria of the study. Thirty-five patients (25%) were excluded because they did not give consent. No patient had respiratory obstruction precipitated by LNR. Nasopharyngeal laryngoscopy was performed in 71 (67.6%) patients, 34 (32.4%) patients were followed-up by telephone interview. (figure 2). The characteristics and symptoms of the 105 enrolled patients are shown in table 1.

    View this table:
    Table 1

    Demographics and clinical symptoms of enrolled patients

    Figure 2
    Figure 2

    Details of the patients who underwent lateral neck radiographs.

    Twenty-nine patients had positive LNR. Interobserver agreement for the thumb sign was substantial, at 92.1% (12.1% agreement on positive ratings and 80.0% agreement on negative ratings), with a κ value of 0.71 (95% CI 0.55 to 0.87). Interobserver agreement for the vallecula sign was substantial, at 92.9% agreement (13.6% agreement on positive ratings and 79.3% agreement on negative ratings), with a κ value of 0.75 (95% CI 0.60 to 0.90). Of the 29 patients with an abnormal radiograph, 26 underwent LNR and 3 were discharged by the emergency physician. Of the 76 patients with a normal radiograph, 45 underwent LNR and the rest were discharged. Among patients who had normal LNR and were discharged with phone follow-up, 71.0% had a sore throat and 41.9% had odynophagia, while among those who had a normal x-ray and laryngoscopy, 100% had a sore throat and 68.9% had odynophagia. In sum, 71 patients underwent laryngoscopy, and supraglottitis was diagnosed in 21 (grade 1=7, grade 2=8, grade 3=6) (table 2), for a total incidence of 20% (figure 2). Supraglottitis was diagnosed in 17 of the 26 (65%) patients who had abnormal x-rays, and four of the patients (13%) who had normal xrays. No patients who were discharged had supraglottitis.

    View this table:
    Table 2

    Results and outcomes of lateral neck radiographs

    Half of the laryngoscopies were performed within 1 h of the LNR. The median time interval from LNR to laryngoscopy was 1 h (mean±SD, 3.6±5.2). Laryngoscopies were delayed in patients who had normal radiographs and were less likely to have supraglottitis and in those who underwent radiographs on the weekend. All of the laryngoscopies were performed within 24 h of the LNR. Antibiotics were prescribed by emergency department physicians in four cases before the subsequent laryngoscopy, while six patients who did not undergo laryngoscopy were treated with antibiotics.

    A radiographic abnormality was seen in 57.1% of the radiographs with grade 1 supraglottitis. Four patients (19%) with supraglottitis had false negative LNR; 3 of these were grade 1. All patients with grade 3 supraglottitis, which is an indication for artificial airway management,15 had an abnormality on x-ray (positive thumb or vallecula sign).; 87.5% (7/8) of grade 2 supraglottitis patients had detectable radiographic abnormalities. (table 2). In all patients who were followed-up by telephone interview, symptoms had resolved.

    The test characteristics of the LNR using the final interpretation agreed on by both the emergency physician and radiologist are displayed in Table 3. The LR positive of the LNR was 5.67 (95% CI 3.27, 9.82) and the LR negative was 0.22 (95% CI 0.10, 0.51). The sensitivity of the emergency physician's interpretation was slightly higher than the radiologist's 81% (62.2, 97.8) vs. 66.7% (95% CI 46.5, 86.8), but specificity was similar (84.5% vs. 85.7%).

    View this table:
    Table 3

    Lateral neck radiograph test characteristics

    Six of the patients with negative LNR who did not undergo larygnoscopy were treated with antibiotics. The diagnoses of these patients were acute pharyngitis (n=4), tonsillitis (n=1) and intestinal perforation (n=1). We could not be certain if these six patients had supraglottitis. If we re-analysed the results without these six patients, specificity decreased from 85.7% to 84.6 (76.6 to 92.6)%.

    Discussion

    In this prospective study, LNR had relatively low sensitivity and specificity for supraglottitis in adults. Previous studies reported a sensitivity and specificity of the classic thumb sign of 83–100% and 87–100%, respectively.6 ,8 In 1997, Ducic et al7 reported a new radiographic sign, vallecula sign, which had 98.2% sensitivity and 99.5% specificity in their report. Based on these findings, some authors suggested that the LNR might obviate the use of routine laryngoscopy. However, these retrospective studies were at high risk of overestimating the diagnostic value, and no validation study of the vallecula sign has been reported.

    The LNR is often used to rule out a diagnosis of supraglottitis; hence a low negative likelihood ratio and high negative predictive value are needed. In this study, the negative likelihood ratio was 0.22, which generates small changes in probability, and the negative predictive value was 94.7%, which indicated that 1 in 20 patients would be missed by LNR. As ruling out supraglottitis based on negative LNR findings might result in misdiagnosis, for patients with probable supraglottitis, nasopharyngeal laryngoscopy should be considered to make a definite diagnosis. However, compared with paediatric supraglottitis, adult supraglottitis is less severe, with a lower risk of airway obstruction, especially for grade 1 supraglottitis. Thus, if the role of LNR is to detect only severe supraglottitis, then perhaps these x-rays have sufficient diagnostic accuracy.

    In this study, one emergency physician and one radiologist interpreted the LNR independently, and we evaluated interobserver disagreement. The thumb and vallecula signs were in agreement about 92% of the time, and the kappa value suggested moderate interobserver agreement. The radiologist is an expert in the interpretation of radiographs, including LNR. However, in Japan, emergency physicians almost always interpret LNR. This is because of the increased need for interpretation of CT, MRI and positron emission tomography images by radiologists, whose time might otherwise be spent interpreting LNR. The sensitivity of the LNR read only by an emergency physician was higher than the radiologist's, but specificity was similar.

    Strengths and limitations

    To our knowledge, this is the first prospective validation study of LNR for detection of adult supraglottitis in a general emergency department population. As previous studies calculating the diagnostic accuracy were limited by spectrum bias,6 ,7 the true sensitivity and specificity were not clear. The incidence of supraglottitis in this study was similar to that in previous reports,12 ,13 so our results can be generalised to other clinical settings. Kurashiki Central Hospital serves 800 000 people in the western area of Okayama prefecture, where few hospitals have more than one otolaryngologist to manage supraglottitis. Thus a patient with suspected supraglottitis at other hospitals in the western Okayama prefecture would be transferred to Kurashiki Central Hospital. In our study, the incidence of supraglottitis was 20.0%. The pre-test probability might change according to the clinical setting and the physician's decision. The negative/positive predictive value of this study will be affected by the incidence in the population. This study included only adults, and it might not be possible to generalise our results to paediatric supraglottitis which may have a more serious course.

    The study had some limitations. Firstly, not all patients underwent laryngoscopy to verify the diagnosis. It is difficult to have an otolaryngologist available at all times to perform laryngoscopy because they tend to have other commitments. If a patient is suspected of having supraglottitis, because of the risk of airway obstruction, laryngoscopy takes priority over the otolaryngologist's other work, while in patients who are suspected of having the common cold, laryngoscopy is not prioritised. It is improper to make the latter patients wait for laryngoscopy and also ethically improper to subject a patient suspected of having a common cold to unnecessary invasive laryngoscopy. Therefore, we defined the reference standard as laryngoscopy or telephone follow-up. In this study, only 57% (4/7) of grade 1 supraglottitis patients had a radiological abnormality, and 8.9% (4/45) of patients with normal radiographs and laryngoscopy had supraglottitis. Patients who had normal radiographs without laryngoscopy were followed-up by telephone interview because the emergency department physician estimated that they had a low probability of supraglottitis. It is possible that a maximum of 8.9% of patients followed-up by telephone had supraglottitis.

    35 patients were excluded because they did not consent to participate in the study. Most of the excluded patients had normal radiographs and a lower rate of odynophagia (65.7% in the enrolled patients vs 42.9% in the excluded patients), and were younger (42.8±18.1 years for the enrolled patients vs 36.1±16.5 for the excluded patients). Patients at low risk of epiglottitis were excluded from the study, which might decrease the validity of the study. In all of the patients who were followed-up by telephone interview, symptoms had resolved, and none was diagnosed with supraglottitis. Therefore, it is unlikely there were many missed patients with supraglottitis and hence missed cases should have had little, if any, effect on our results.

    We enrolled only 21 patients with supraglottitis, similar to previous reports.6 ,7 The small sample size resulted in wide 95% CIs, and further research with larger samples is needed to calculate the test accuracy precisely.

    Finally, laryngoscopy was delayed for a maximum of 24 h after LNR; while half of the laryngoscopies were performed within 1 h of the LNR, a delay could have resulted in more negative laryngoscopies.

    Conclusion

    LNR in our practice had a 0.22 negative likelihood ratio for supraglotittis. Although the sensitivity of LNR was 100% for severe (grade 3) supraglottitis, the test the test cannot be used to rule out mild supraglottitis when the pre-test probability is high. Physicians should consider nasopharyngeal laryngoscopy in patients with suspected supraglottitis, even if there are no radiographic abnormalities.

    References

    1. ↵
      1. Rafei K,
      2. Lichenstein R
      . Airway infectious disease emergencies. Pediatr Clin North Am 2006;53:215–42.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Sheikh KH,
      2. Mostow SR
      . Epiglottitis—an increasing problem for adults. Western J Med 1989;151:520–4.
      OpenUrlPubMedWeb of Science
    3. ↵
      1. Harris RD,
      2. Berdon WE,
      3. Baker DH
      . Roentgen diagnosis of acute epiglottitis in the adult. J Can Assoc Radiol 1970;21:270–2.
      OpenUrlPubMed
    4. ↵
      1. Schabel SI,
      2. Katzberg RW,
      3. Burgener FA
      . Acute inflammation of epiglottitis and supraglottic structures in adults. Radiology 1977;122:601–4.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Schumaker HM,
      2. Doris PE,
      3. Birnbaum G
      . Radiographic parameters in adult epiglottitis. Ann Emerg Med 1984;13:588–90.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Rothrock SG,
      2. Pignatiello GA,
      3. Howard RM
      . Radiologic diagnosis of epiglottitis: objective criteria for all ages. Ann Emerg Med 1990;19:978–82.
      OpenUrlCrossRefPubMed
    7. ↵
      1. Ducic Y,
      2. Hebert PC,
      3. MacLachlan L,
      4. et al
      . Description and evaluation of the vallecula sign: a new radiologic sign in the diagnosis of adult epiglottitis. Ann Emerg Med 1997;30:1–6.
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Yong MG,
      2. Choo MJ,
      3. Yum CS,
      4. et al
      . Radiologic laryngeal parameters in acute supraglottitis in Korean adults. Yonsei Med J 2001;42:367–70.
      OpenUrlCrossRefPubMed
    9. ↵
      1. Frumkin K
      . Vallecula sign and epiglottitis. Ann Emerg Med 1998;31:139.
      OpenUrlPubMed
    10. ↵
      1. Jaeschke R,
      2. Guyatt G,
      3. Sackett DL
      . Users’ guides to the medical literature. III. How to use an article about a diagnostic test. A. Are the results of the study valid? Evidence-Based Medicine Working Group. JAMA 1994;271:389–91.
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Bossuyt PM,
      2. Reitsma JB,
      3. Bruns DE,
      4. et al
      . Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. BMJ 2003;326:41–4.
      OpenUrlAbstract/FREE Full Text
    12. ↵
      1. Briem B,
      2. Thorvardsson O,
      3. Petersen H
      . Acute epiglottitis in Iceland 1983–2005. Auris Nasus Larynx 2009;36:46–52.
      OpenUrlCrossRefPubMedWeb of Science
    13. ↵
      1. Guldfred LA,
      2. Lyhne D,
      3. Becker BC
      . Acute epiglottitis: epidemiology, clinical presentation, management and outcome. J Laryngol Otol 2008;122:818–23.
      OpenUrlPubMed
    14. ↵
      1. Katori H,
      2. Tsukuda M
      . Acute epiglottitis: analysis of factors associated with airway intervention. J Laryngol Otol 2005;119:967–72.
      OpenUrlPubMedWeb of Science
    15. ↵
      1. Frantz TD,
      2. Rasgon BM,
      3. Quesenberry CP Jr.
      . Acute epiglottitis in adults. Analysis of 129 cases. JAMA 1994;272:1358–60.
      OpenUrlCrossRefPubMedWeb of Science

    Supplementary materials

    • Abstract in Japanese

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Contributors TaF and ToF designed the study protocol. HO and YO interpreted the lateral neck radiographic data (index test). KI and TaF interpreted the nasopharyngeal laryngoscopy data (reference standard test). TaF was primarily responsible for writing the draft. All authors revised and approved submission of the current manuscript.

    • Competing interests None.

    • Ethics approval The study was approved by the institution review board of Kurashiki Central Hospital.

    • Provenance and peer review Not commissioned; externally peer reviewed.