Program in Epidemiology, Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109, USA.
Clin Cancer Res. 2013 Mar 1;19(5):1197-203. doi: 10.1158/1078-0432.CCR-12-2647. Epub 2013 Jan 14.
To identify a prognostic gene signature for patients with human papilloma virus (HPV)-negative oral squamous cell carcinomas (OSCC).
Two gene expression datasets were used: a training dataset from the Fred Hutchinson Cancer Research Center (FHCRC, Seattle, WA; n = 97) and a validation dataset from the MD Anderson Cancer Center (MDACC, Houston, TX; n = 71). We applied L1/L2-penalized Cox regression models to the FHCRC data on the 131-gene signature previously identified to be prognostic in patients with OSCCs to identify a prognostic model specific for patients with high-risk HPV-negative OSCCs. The models were tested with the MDACC dataset using a receiver operating characteristic (ROC) analysis.
A 13-gene model was identified as the best predictor of HPV-negative OSCC-specific survival in the training dataset. The risk score for each patient in the validation dataset was calculated from this model and dichotomized at the median. The estimated 2-year mortality (± SE) of patients with high-risk scores was 47.1% (± 9.24%) compared with 6.35% (± 4.42) for patients with low-risk scores. ROC analyses showed that the areas under the curve for the age, gender, and treatment modality-adjusted models with risk score [0.78; 95% confidence interval (CI), 0.74-0.86] and risk score plus tumor stage (0.79; 95% CI, 0.75-0.87) were substantially higher than for the model with tumor stage (0.54; 95% CI, 0.48-0.62).
We identified and validated a 13-gene signature that is considerably better than tumor stage in predicting survival of patients with HPV-negative OSCCs. Further evaluation of this gene signature as a prognostic marker in other populations of patients with HPV-negative OSCC is warranted.
鉴定人乳头瘤病毒（HPV）阴性口腔鳞状细胞癌（OSCC）患者的预后基因特征。
使用了两个基因表达数据集：弗雷德·哈钦森癌症研究中心（西雅图，华盛顿州；n = 97）的训练数据集和 MD 安德森癌症中心（休斯顿，得克萨斯州；n = 71）的验证数据集。我们将 L1/L2-惩罚 Cox 回归模型应用于之前在 OSCC 患者中鉴定为具有预后意义的 131 个基因特征的 FHCRC 数据，以确定特定于 HPV 阴性 OSCC 高危患者的预后模型。使用 MDACC 数据集通过接收者操作特征（ROC）分析对模型进行测试。
在训练数据集中，确定了一个 13 基因模型，作为 HPV 阴性 OSCC 特异性生存的最佳预测因子。验证数据集中每位患者的风险评分均根据该模型计算，并以中位数进行二分类。高危评分患者的估计 2 年死亡率（± SE）为 47.1%（± 9.24%），而低危评分患者为 6.35%（± 4.42%）。ROC 分析显示，风险评分[0.78；95%置信区间（CI），0.74-0.86]和风险评分加肿瘤分期（0.79；95%CI，0.75-0.87）的年龄、性别和治疗方式调整模型的曲线下面积明显高于仅基于肿瘤分期（0.54；95%CI，0.48-0.62）的模型。
我们鉴定并验证了一个 13 基因特征，其在预测 HPV 阴性 OSCC 患者的生存方面明显优于肿瘤分期。需要进一步评估该基因特征作为其他 HPV 阴性 OSCC 患者人群的预后标志物。
