2018-19 Section 7-Neoplastic and Inflammatory Diseases of the Head and Neck eBook

First page Previous page 26 Next page Last page

Original Article

TABLE 1. Continued

Univariate Survival Analysis

Multivariable Survival Analysis

Characteristics

95% CI

Surgical margin c Negative

1.000 1.827

1.000 1.261

Positive

1.651

2.022

< .001

1.134

1.402

< .001

Postoperative RT No

1.000 1.228

Yes

1.098

1.374

< .001

Postoperative chemotherapy No

1.000 1.950

1.000 1.161

Yes

1.734

2.193

< .001

1.017

1.324

.027

Abbreviations: 95% CI, 95% confidence interval; ENE, extranodal extension; HR, hazard ratio; LN, lymph node; OS, overall survival; RT, radiotherapy. Final multivariable models were determined after stepwise backwards selection. a Number of positive metastatic LNs was modeled with 4 knots at the 63rd, 71st, 83rd, and 98th quantiles (1, 2, 6, and 37 LNs, respectively) and the number of LNs examined was modeled with 3 knots at the 10th, 50th, and 90th quantiles (11, 26, and 52 LNs, respectively). b Variables dropped out of the model. c Missing data were imputed by multiple imputation. d Multivariable model was adjusted for postoperative RT and tumor grade by stratification due to nonproportional hazards.

2324 patients with LN-negative disease (see Support- ing Table 1). The median follow-up was 54.3 months. The mean number of LNs examined was 28.1 ( 6 the standard deviation [SD] of 16.3), and the mean number of metastatic LN identified was 4.2 ( 6 SD of 10.2). In total, 78.9% (1733 patients) and 57.5% (1337 patients), respectively, of patients with LN-positive and LN-negative disease underwent adju- vant radiotherapy( P < .001). Similarly, 30.4% (667 patients) of patients with LN-positive disease and 5.3% (123 patients) of patients with LN-negative dis- ease received adjuvant chemotherapy ( P < .001). The percentage of patients with positive LNs varied sub- stantially by histology, with positive LNs found in approximately 75% of salivary duct carcinoma, 65% of adenocarcinoma, 55% of other carcinomas (includ- ing carcinoma ex pleomorphic adenoma), 41% of mucoepidemoid carcinoma, 34% of adenoid cystic carcinoma, and 26% of acinic cell carcinoma neck dissection specimens, respectively. Number of Metastatic LN In univariate analysis, an increasing number of metastatic LNs was found to be strongly predictive of worse OS ( P < .001) (Table 1). The estimated 5-year OS rates were 81.7%, 60.6%, 36.9%, 30.1%, and 13.9%, respectively, for patients with 0, 1 to 2, 3 to 9, 10 to 19, and 20 meta- static LNs (see Supporting Fig. 2A). A similar impact for the number of metastatic LNs was seen in the patients in the N2b subgroup (see Supporting Fig. 2B). After adjust- ing for other clinical and demographic factors using mul- tivariable Cox regression (Table 1), the number of metastatic LNs remained strongly associated with OS

OS using restricted cubic splines. The optimal number of knots was selected based on the Akaike information crite- rion. Knot locations were placed in default quantiles as described by Harrell. 21 For the number of positive LNs, 4 knots were placed at the 5th, 35th, 65th, and 95th percen- tiles, respectively, corresponding to 1, 2, 6, and 37 LN 1 . For the number of LNs examined, 3 knots were placed at the 10th, 50th, and 90th percentiles, respectively, corre- sponding to 11, 26, and 52 LN 1 . Change points were identified by fitting a piecewise linear regression model on the log relative hazard of the number of LN 1 and the number of LNs examined. 22,23 Recursive partitioning analysis (RPA) with a condi- tional inference tree was used to develop a novel LN stag- ing system. The conditional inference tree was created using independent LN predictors of mortality, and esti- mated by binary recursive partitioning in a conditional inference framework developed by Hothorn et al, Hothorn and Zeileis, and Strasser and Weber. 24-26 Performance of the RPA-derived LN staging system was compared with the AJCC eighth edition staging system using c-indices in patients with determinable AJCC stage. Internal validation was performed using bootstrapping with 1000 replicates to correct for possible optimism in c-indices. All statistical analyses were performed with R (ver- sion 3.4.0; R Foundation, Vienna, Austria) with 2-sided tests and a significance level of .05.

RESULTS Patient Cohort

Overall, 4520 patients met the inclusion criteria, including 2196 patients with LN-positive disease and

Cancer

Made with FlippingBook Annual report