Incremental Prognostic Value of Apparent Diffusion Coefficient Histogram Analysis in Head and Neck Squamous Cell Carcinoma

Abstract

We aimed to investigate the incremental prognostic value of apparent diffusion coefficient (ADC) histogram analysis in patients with head and neck squamous cell carcinoma (HNSCC) and integrate it into a multivariate prognostic model. A retrospective review of magnetic resonance imaging findings was conducted in patients with pathologically confirmed HNSCC between June 2012 and December 2015. For each tumor, six histogram parameters were derived: the 10th, 50th, and 90th percentiles of ADC (ADC10, ADC50, and ADC90); mean ADC values (ADCmean); kurtosis; and skewness. The clinical variables included age, sex, smoking status, tumor volume, and tumor node metastasis stage. The association of these histogram and clinical variables with overall survival (OS) was determined. Further validation of the histogram parameters as independent biomarkers was performed using multivariate Cox proportional hazard models combined with clinical variables, which was compared to the clinical model. Models were assessed with C index and receiver operating characteristic curve analyses for the 12- and 36-month OS. Ninety-six patients were eligible for analysis. Median follow-up was 877 days (range, 54-1516 days). A total of 29 patients died during follow-up (30%). Patients with higher ADC values (ADC10 > 0.958 × 10-3 mm2/s, ADC50 > 1.089 × 10-3 mm2/s, ADC90 > 1.152 × 10-3 mm2/s, ADCmean > 1.047 × 10-3 mm2/s) and lower kurtosis (≤0.967) were significant predictors of poor OS (P < .100 for all). After adjusting for sex and tumor node metastasis stage, the ADC90 and kurtosis are both significant predictors of OS with hazard ratios = 1.00 (95% confidence interval: 1.001-1.004) and 0.58 (95% confidence interval: 0.37-0.90), respectively. By adding the ADC parameters into the clinical model, the C index and diagnostic accuracies for the 12- and 36-month OS showed significant improvement. ADC histogram analysis has incremental prognostic value in patients with HNSCC and increases the performance of a multivariable prognostic model in addition to clinical variables.