Refining the head and neck cancer referral guidelines: a two centre analysis of 4715 referrals

Abstract

Our aim was to identify the set of referral criteria that will offer optimal diagnostic efficacy in patients suspected to have head and neck cancer (HNC) in the primary care setting. We analysed the referral criteria and outcomes from two tertiary care cancer centres in the United Kingdom. Between 2007 and 2010, 4715 patients were referred via the fast track system with a suspected HNC. The main outcome measures were the parameters of diagnostic efficacy, a multivariate regression model to calculate estimated probability of HNC and the area under the receiver operating characteristic curve (AUROC). We found that the majority of referring symptoms had a positive predictive value higher than the 3% cut-off point stated to be significant for HNC detection in the 2015 NICE recommendations. Nevertheless, our multivariate analysis identified 9 symptoms to be linked with HNC. Of these, only 4 are included in the latest NICE guidelines. The best fit predictive model for this dataset included the following symptoms: hoarseness>3 weeks, dysphagia>3 weeks, odynophagia, unexplained neck mass, oral swelling >3 weeks, oral ulcer >3weeks, prolonged otalgia with normal otoscopy, presence of blood in mouth with concurrent sensation of lump in throat, and presence of otalgia with concurrent lump in throat sensation. Intermittent hoarseness and sensation of lump in throat were negatively associated with HNC. The AUROC demonstrated that our model had a higher predictive value (0.77) compared to those generated using the NICE 2005 (0.69) and 2015 (0.68) referral criteria (p<0.0001). An online risk calculator based on this study is available at http://www.orlhealth.com/risk-calculator.html. This paper presents a significantly refined version of referral guidelines which demonstrate greater diagnostic efficacy than the current NICE guidelines. We recommend that further iterative refinements of referral criteria be considered when referring patients with suspected HNC.

External validation of a cancer risk prediction model for suspected head and neck cancer referrals

“…This symptom was the most strongly associated with a cancer diagnosis yielding a sensitivity of 56.9% followed by persistent hoarseness for more than 3 weeks (Table ). The presenting symptoms have been grouped similarly to the previous study to allow direct comparisons …”

“…The latest iteration of the NICE guidelines published in 2015 did not manage to increase the detection rate of cancer via the urgent suspicion of cancer pathway . When the 2015 NICE symptom list was modelled a low HNC predictive power was found . A large number of patients are seen urgently but are at low risk of having HNC.…”

“…Head and neck cancer probability for each individual patient was calculated using the symptoms on the published HNC risk calculator. Patient's age and gender were also significant predictors of HNC and were included in the calculator matrix . The sensitivity, specificity and c‐statistic of the calculator based on the Scottish cohort of patients are presented and compared to the previously published data set.…”

“…The sensitivity, specificity and c‐statistic of the calculator based on the Scottish cohort of patients are presented and compared to the previously published data set. The same probability cut‐off was applied, set at 0.08 . Chi‐square test and independent t test were performed for univariate assessment of association for categorical and continuous variables, respectively.…”

“…To refine the referral guidelines, a retrospective cohort study was undertaken to examine the symptoms related to HNC. A risk calculator was generated from this study using a combination of symptoms that had a statistically significant higher predictive value when compared to the NICE recommendations . The list of significant symptoms is presented in Table .…”